use of ground and air-based photogrammetry for soil erosion … · 2019. 9. 30. · the water...
TRANSCRIPT
Use of ground and air-based photogrammetry
for soil erosion assessment
Bernardo Moreira Candido
This dissertation is submitted for the degree of Doctor of
Philosophy
September 2019
Lancaster Environment Centre
I dedicate this thesis to my dear cousin, friend and brother
Clebinho (In Memorian), for being part of the most important
moments of my life.
May God always bless you.
Declaration
This thesis has not been submitted in support of an application for another
degree at this university. It is the result of my own work and includes nothing
that is the outcome of work done in collaboration except where specifically
indicated. Many of the ideas in this thesis were the product of discussion with
my supervisors John N. Quinton, Michael R. James and Marx L. N. Silva.
Bernardo Moreira Candido
Lancaster University, UK
ACKNOWLEDGMENTS
This thesis would not have been possible without the support and guidance of
many people. I would like to thank:
My wife Isabella, for doing this PhD with me. The many challenges brought by
this PhD were overcome by both of us, together.
My parents, who taught me responsibility and respect. My grandparents, who
gave me all the support to live and study in Lavras.
My brothers, Vitor Hugo, Pablo and Miguel, for always remind me of smiling.
I would particularly like to thank my supervisors for their constant support,
advices and patience over the past few years:
Professor Marx Silva, who believed in me since the beginning and always
supported me to dedicate myself to one of the things that I most love to do,
learning.
Professor John Quinton, who I am glad to have had the opportunity of meeting
and knowing his inspiring way of doing science and dealing with those starting
in the sciences, like me.
Professor Mike James, who made me even more fascinated with
photogrammetry through his papers and through the interesting discussions we
had while I joined his group.
Special mention must go to Wellington de Lima, for his assistance in setting up
the flights over the experiments at UFLA while I was in Lancaster.
I thank the society from where the resources for my research came from. I thank
the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq,
grant numbers 233547/2014-2, 149622/2014-7, 306511/2017-7 and 202938/2018-2)
for providing the scholarship and all necessary resources while I was in
Lancaster, UK. I also thank the Fundação de Amparo a Pesquisa do Estado de
Minas Gerais (FAPEMIG, grant numbers APQ-00802-18 and CAG-APQ 01053-
15) for providing the resources to buy equipment and software used in this work.
All my colleagues from the Federal University of Lavras for a very nice
environment where I have been throughout the last 10 years.
All my friends and brother from UDV, for being my second family while I was
in Lavras.
Lancaster University and the Lancaster Environment Centre. The Federal
University of Lavras and the Department of Soil Science.
Finally, I am grateful to God and M. Gabriel, who taught me the importance of
being humble and gave me the strength to overcome the challenges of this PhD.
GENERAL ABSTRACT
Water erosion affects all types of soils around the world at different intensities.
However, in tropics water erosion is the most important form of soil erosion and
has received much concern in the last decades. The major challenge in soil
conservation is the development and implementation of strategies to mitigate the
erosion processes in urban and rural areas. Thus, understanding the processes
involved in each type of water erosion (sheet, rill and gully erosion), as well as
its quantification, is a key factor in managing and developing soil conservation
and erosion mitigation strategies. In that way, this thesis aims to investigate the
efficiency of ground and air-based photogrammetry for soil erosion assessment,
as well as to address some gaps in our understanding of the evolution of erosive
processes in its different forms of occurrence. In doing so, we evaluated the
factors that influence the development of erosion in micro and macro scales, with
experiments in the laboratory and in the field. In the first chapter, it was
evaluated the influence of gradient change and runoff volumes on rill erosion
process, using digital close-range photogrammetry in a laboratory soil flume. In
addition, morphological rill parameters were estimated to allow a better
understanding of the rill erosion behaviour under different treatments. The
results showed that the flow velocity in rills increased with water flow and slope,
showing a strong correlation with the amount of rill erosion. On steep slopes the
soil erosion was dominated by the rill erosion with less rill network density
while, on low slopes, there were other types of soil erosion occurring together
with rill erosion, causing the reduction of soil loss due to rill erosion. The digital
close-range photogrammetry technique provided millimetric precision, which is
sufficient for rill erosion investigations. In the second chapter aimed to evaluate
the efficiency of SfM based on UAV images in obtaining accurate measurements
of soil loss in areas of sheet erosion, under natural rainfall, where channelized
erosion is not the principal mechanism. The measurements acquired from SfM
were compared to the sediments collected in each soil erosion plots. The results
of the soil losses obtained by UAV-SfM presented a high correlation with the
sediments collected in the plots. This is of great relevance in the context of the
monitoring and modelling of water erosion, since the quantification of soil loss
around the world is mainly done using plots, a method that presents high
operational cost. In addition, the study of laminar erosion through the UAV-SfM
allows not only to calculate the soil loss values but to visualize the spatial
variation of the erosion process (detachment, transport and deposition)
practically in real time along the area. In the third chapter it was evaluated the
application of UAV-SfM technique in a gully system. For the first time, a study
was carried out evaluating the relative contribution of the different types of
erosion (sheet, rill and gully sidewall) in the gully development. This was
possible due to the millimetric level of precision of the point clouds, allowing to
evaluate even the contribution of the laminar erosion, which is new in gullies
studies. As a result, it was possible to quantify sediments volumes stored in the
channels and lost from the gully system, as well as to determine the main
sediment sources. The study suggests that the main source of sediments in the
gully was due to the mass movements, followed by rills and sheet erosion. The
UAV-SfM proved to be effective in the gully monitoring. The results findings by
this thesis indicate that the use of ground and air-based photogrammetry are
precise tools in detecting soil surface changes and can be used to assess water
erosion in its various forms of occurrence in nature. In addition, the UAV-SfM
has proven to be a very useful technique for monitoring soil erosion over time,
especially in hard-to-reach areas.
Keywords: Structure-from-motion. Unmanned aerial vehicle. Digital close-range
photogrammetry. Erosion plot. Sheet erosion. Rill erosion. Gully erosion.
Contents
1. LITERATURE REVIEW ........................................................................................... 1
1.1 Soil erosion ............................................................................................................ 1
1.2 Soil erosion by water ........................................................................................... 2
1.2.1 Sheet erosion .................................................................................................. 4
1.2.2 Rill erosion ..................................................................................................... 7
1.2.3 Gully erosion .................................................................................................. 9
1.3 Topographic methods for soil erosion assessment ....................................... 11
1.3.1 High-resolution topographic methods .................................................... 12
1.3.2 Digital photogrammetry ............................................................................ 13
1.3.3 Unmanned aerial vehicles (UAV) ............................................................. 14
1.3.4 Structure-from-Motion (SfM) .................................................................... 15
1.4 Thesis aims and objectives ................................................................................ 17
1.5 References ............................................................................................................ 19
2. RILL EROSION AND MORPHOLOGY DEVELOPMENT IN RESPONSE
TO CHANGING DISCHARGE AND SLOPE PROFILES .................................. 31
3. HIGH RESOLUTION MONITORING OF SHEET (INTERRILL) EROSION
USING STRUCTURE-FROM-MOTION................................................................ 61
4. EVALUATION OF SEDIMENT SOURCE AND VOLUME OF SOIL
EROSION IN A GULLY SYSTEM USING STRUCTURE-FROM-MOTION
AND UAV DATA: A CASE STUDY FROM MINAS GERAIS, BRAZIL ........ 88
5. GENERAL CONCLUSIONS ............................................................................... 111
6. FUTURE WORK .................................................................................................... 113
1
1. Literature Review
1.1 Soil erosion
Soil is a non-renewable resource and essential to the maintenance of
humans and natural ecosystems, ensuring food, fibre, energy, and water quality.
However, the unplanned use of soil can cause its degradation, with negative
economic and social impacts on present and future generations. Currently, about
800 million people are food insecure around the world (Lal, 2013) and 2 billion
have no access to safe and affordable water (World Health Organization and
UNICEF, 2015; Paul Obade and Lal, 2016). In addition, the world's population
will probably reach 9·1 billion people in 2050 and, consequently, the demand for
food is also expected to increase dramatically (Alexandratos and Bruinsma,
2012).
Considering that about 50 % of the earth's surface is dedicated to
agriculture and more than 99·7% of the human’s food comes from the land
(Pimentel, 2006) the soil is a key factor in order to provide food and water security
for future generations. Thus, understanding the factors that cause its degradation
is essential to implement reclamations and preservations measures of ecosystems
and rural areas.
In natural conditions, the slow loss of soil by the erosion process has been
responsible for transforming and sculpting the landscape throughout geologic
2
time, through the mobilisation and deposition of soil particles, by water and air.
However, whilst the natural soil erosion is a feature of any terrestrial ecosystem,
anthropogenic soil erosion accelerates the geomorphic processes of natural soil
erosion (Lal, 2001).
Anthropogenic soil erosion, most associated with inappropriate land use,
has significantly accelerated the rate of soil loss, causing considerable impacts in
environment, economy and society worldwide. According to Pimentel (2006),
about 80% of the agricultural land around the world suffers from severe or
moderate soil erosion and 10% shows slight signals of erosion. As a result, about
30% of productive areas worldwide have become infertile and abandoned for
agriculture use (Pimentel, 2006). Therefore, due to the long-term negative impact
on the soil properties, soil erosion becomes the most important land degradation
problem in the world (Eswaran et al., 2001) and has been prioritized by
environmental scientists and policymakers worldwide.
1.2 Soil erosion by water
Water erosion affects all types of soils around the world at different
intensities (Blanco and Lal, 2010). However, in tropics water erosion is the most
important form of soil erosion and has received much concern in the last decades.
It is one of the most serious forms of soil degradation and a major threat to the
sustainability of agroecosystems, affecting around one billion hectares
worldwide (Lal, 2003; Luffman et al., 2015; Xu et al., 2016; Guo et al., 2017).
3
The water erosion is based on the processes of detachment, transport and
deposition of soil particles by the impact of raindrop and overland runoff, mainly
taking the form of concentrated and dispersed flow (Ellison, 1947; Kinnell, 2006;
Shi et al., 2010). This refers to the movement of soil particles along the surface,
with deposition of eroded materials in the lower slope regions and in
watercourses (Horton, 1945). This eroded material may generate new soil, or
simply silting rivers and lakes. In addition, these sediment losses are often
associated with organic matter, nutrients, heavy metals and other contaminants,
resulting in direct and indirect impacts on soil functions, and on aquatic and
terrestrial ecosystems (Glendell and Brazier, 2014; Rickson, 2014).
Different factors influence the evolution of the erosive process, such as the
intensity and duration of rainfall, soil properties, slope, vegetation cover and soil
surface roughness (Le Bissonnais et al., 2005; Mahmoodabadi and Sajjadi, 2016;
Wang et al., 2017; Hao et al., 2019).
The impact of raindrops plays an important role in water erosion because,
in addition to promoting the detachment of soil particles, it also promotes
increased transport of sediment through runoff (Zhang and Wang, 2017) (Figure
1). Soil properties, such as the texture and aggregates stability, affect soil
resistance to detachment and water infiltration (Le Bissonnais, 1996). Thus, the
effects of rain intensity on erosion strongly depends on soil type (Defersha and
Melesse, 2012; Wu et al., 2018).
4
Fig. 1. Illustration of rainsplash erosion and soil-particle mobilization, transport,
and deposition.
The major challenge of soil conservation today is the efficient
implementation of conservation management practices in agricultural and urban
areas. For this, the knowledge of the dynamics of erosion in the area, through the
spatiotemporal studies of erosive processes and types of erosion (sheet, rills and
gullies), as well as its quantification, are key factors in managing and developing
soil conservation and erosion mitigation strategies.
1.2.1 Sheet erosion
Sheet erosion, also known as laminar and interrill erosion, mainly affects
agricultural areas in sloping regions in tropical environments (Le Bissonnais et
5
al., 1998; Sirjani and Mahoodabadi, 2014). Once it is started, the runoff develops
in small rills and the portion of the overland flow flowing between the rills is
called sheet erosion or interrill erosion. This type of erosion is characterized by
the detachment of fine soil particles present on the surface due to the impact of
raindrops and the dispersed flow over the surface (Kinnell, 2013; Maïga-Yaleu et
al., 2015).
Laminar erosion is the most common type of erosion, corresponding to
about 70% of total soil erosion, occurring simultaneously to other erosive
processes (Blanco and Lal, 2010). This type of erosion is affected by several factors
such as rainfall intensity, slope gradient, vegetation, soil properties and surface
characteristics, such as roughness and crusting (Dlamini et al., 2011; Ibáñez et al.,
2016; Zhang et al., 2019).
During the sheet erosion process, a mixture of water and solid particles
flows over the soil as a sheet, causing erosion through loss of successive layers of
soil (Fournier, 1960). Sheet erosion can carry only fine particles, such as clay and
silt, while coarse particles are associated with the development of rill erosion
(Hao et al., 2019). This selective behaviour in the transport of the most useful soil
particles is a harmful aspect of the sheet flow. Thus, in hillslopes sheet erosion is
much more severe than gully erosion (Descroix et al., 2008), contributing to soil
fertility loss, reduced productivity and being the initial stage of rill erosion
(Müller-Nedebock et al., 2016).
The correct measurement of soil losses from sheet erosion is a key factor
for the development of conservation planning. However, the main form of
6
monitoring and quantification of sheet erosion in the world is through erosion
plots under natural and artificial rainfall (Cerdan et al., 2010; García-Ruiz et al.,
2015; Guo et al., 2015; Zhao et al., 2019). This traditional method is considered a
time-consuming and costly process, restricting the evaluation of sheet erosion
only in places where there is a plot system installed in the field (Cerdan et al.,
2010).
With the recent advances in technology, the use of high-resolution digital
elevation models (DEM) has assumed an important role in the study of processes
that occur on the earth's surface. These advances have become possible with the
development of techniques such as Structure-from-Motion (SfM), which
combines the principles of photogrammetry and modern methods of computer
vision (Westoby et al., 2012).
The SfM photogrammetry has been widely used to evaluate erosion in rills
and gullies (Bazzoffi, 2015; Nouwakpo et al., 2015; Hänsel et al., 2016; Neugirg et
al., 2016; Morgan et al., 2017). However, there are no studies with the application
of SfM in the measurement of sheet erosion in areas where there is no incision of
channels of concentrated water flow. Thus, because of the difficulty in
monitoring the loss of millimetre layers of soil, remote monitoring of sheet
erosion is a major challenge in soil science.
7
1.2.2 Rill erosion
Rills are small concentrated flow paths in which the depth and flow
velocity are greater than in the surface runoff and with erosive potential higher
than sheet erosion (Govers et al., 2007; Blanco and Lal, 2010). Rill erosion occurs
from soil detachment by scouring and transport of soil particles by the
concentrated water flow in these narrow channels (Bagnold, 1966).
Rill erosion is the main source of sediments and mechanisms for sediment
transport in erosive processes on hillslopes and is related to the increase in
erosion rates (Bryan, 1990; Rauws, 1987; Di Stefano et al., 2017). Once the rills
established in an area, sediment production can increase by up to 40 times
(Morgan, 1977). In addition, the presence of rills in sloping areas may correspond
to 70% to 90% of total erosion (Renard et al., 1997; Zheng and Tang, 1997).
Rill erosion is widespread mainly in dry regions, being one of the main
forms of soil loss in agricultural lands around the world (Cerdan et al., 2002;
Polyakov and Nearing, 2003; Hancock et al., 2008; Wirtz et al., 2012; Di Stefano et
al., 2013; Jiang et al., 2018; Zhang et al., 2019). In addition, the transport of
sediments from the hillslopes to the watercourses becomes more efficient as a rill
network develops (Berger et al., 2010; Heras et al., 2011; Aksoy et al., 2013; He et
al., 2016; Sofia et al., 2017). Thus, rill erosion affects significantly the
micromorphology of the hillslopes, generating large rill networks in the area and
increasing the connectivity between the sediments, besides promoting
8
permanent changes in the landscape with the evolution of rills to gullies (Berger
et al., 2010; Wirtz et al., 2010).
Several experiments have been conducted to study the formation and
development of rill erosion (Bryan and Rockwell, 1998; Berger et al., 2010; Wirtz
et al., 2012; He et al., 2016; Di Stefano et al., 2017; Lu et al., 2017; Sofia et al., 2017;
Jiang et al., 2018; Zhang et al., 2019). Researches have shown that rill erosion is
directly related to increased runoff (Mancilla et al., 2005), rainfall intensity
(Brunton and Bryan, 2000) and slope (Berger et al., 2010).
However, Wirtz et al. (2012) have shown that rill erosion is not a simple
function related to rainfall intensity or slope gradient, but a complex process that
is influenced by surface sealing, rill development, headcut incision, and sidewall
collapse. Therefore, there are differences in the experimental results obtained,
which vary according to the type of soil, rainfall conditions and spatial and
temporal scales (Devente and Poesen, 2005; Govers et al., 2007).
In the field, the slope gradient is constantly changing along the hillslope,
influencing the flow velocity and, consequently, the rill network. However, due
to the absence of adequate techniques for spatial monitoring of rill development,
few studies are evaluating the influence of gradient change and runoff volume
on the formation of the rill network. Thus, there is a gap to be explored in the
study of morphological indicators and the spatial behaviour of the formation of
rill networks at different slopes and water flow intensities. The understanding of
the evolution and behaviour of rill erosion in each situation helps the
9
development of prevention and recovery strategies in areas affected by this type
of soil erosion.
1.2.3 Gully erosion
Gully erosion is the most severe type of water erosion when compared to
sheet and rill erosion, promoting soil degradation and causing large volumes of
soil loss in agricultural areas (Valentin et al., 2005). Gully erosion occurs when
the concentrated runoff begins to develop deep channels that expand into deep
trenches over time (Luffman et al., 2015). Erosion caused by the concentrated
water flow reduces the soil surface layer through the formation of gullies,
impacting soil fertility and polluting adjacent streams (Allen et al., 2018; Bastola
et al., 2018; Zabihi et al., 2018). In the early stages, the gullies develop rapidly to
great depths, making difficult to control through agricultural machinery and
making the area recovery expensive (Nachtergaele and Poesen, 2002;
Vanwalleghem et al., 2005).
Gullies represent a serious environmental problem in the world, occurring
mainly in arid and semi-arid regions, where vegetation density is low, reducing
soil protection to rainfall and runoff (Sankey and Draut, 2014). Although the
gullies occupy around 5% of the area of a catchment, gully erosion represents the
largest source of soil eroded sediments and may reach 94% of total erosion
(Makanzu Imwangana et al., 2014; Ionita et al., 2015; Bollati et al., 2016). The
development of gullies in an area promotes connectivity between sediment
10
channels and increases the volume of runoff on the hillslope, generating flood
risks and sedimentation of rivers, lakes and reservoirs (Poesen et al., 2003).
Gully erosion is controlled by several factors such as surface runoff, water
movements in the sub-surface, and soil piping (Kirkby and Bracken, 2009; Poesen
et al., 2018). Although the gullies are the result of natural processes that occur on
the soil surface, human action can accelerate the formation and development of
this type of erosion (Thorburn and Wilkinson, 2013; Rodrigo Comino et al., 2015).
Therefore, to prevent these negative effects and implement reclamation
strategies, it is necessary to understand the magnitude of the problem and the
factors causing it (Mitas and Mitasova, 1998; Poesen et al., 2003). Measuring gully
erosion can be very challenging because it is a highly complex process, with
erosion and deposition occurring at the same time and in the same place (Gómez-
Gutiérrez et al., 2014). Consequently, the measurement accuracy of gullies is
directly influenced by the spatial scales and by their morphology (Castillo et al.,
2012). In this way, to provide reliable data about soil losses, several techniques
for monitoring and quantifying gully erosion are currently available, such as pins
(Chaplot, 2013), Light Detection and Ranging (LiDAR) (Perroy et al., 2010), hand-
held mobile laser scanner (James and Quinton, 2014) and aerial photographs and
photogrammetric techniques (Marzolff and Poesen, 2009; Castillo et al., 2012).
Among the new methodologies used to monitor gully erosion,
photogrammetry combined with three-dimensional (3-D) soil surface
reconstruction methods have been widely used recently (Castillo et al., 2012;
Gómez-Gutiérrez et al., 2014; Kaiser et al., 2014; Di Stefano et al., 2017; Ben
11
Slimane et al., 2018). However, although many studies have described the
formation and development of gullies (Harvey, 1992; Vandekerckhove et al.,
1998; Sidorchuk et al., 2003; Conoscenti et al., 2014), few studies have evaluated
the dynamics of the sediments movements and the relative contribution of each
process to total erosion within the gullies complex. The understanding of the
contribution rates of sheet erosion, rills and gully sidewalls, as well as the
quantification of the sediments stored in the channels, is of great importance in
the establishment of strategies for prevention, monitoring and recovery of gullies
(Hosseinalizadeh et al., 2019).
1.3 Topographic methods for soil erosion assessment
During an erosive event, the soil surface is in continuous transformation.
Depending on the volume of soil transported, the erosion processes can result in
considerable topographic variations and may have negative impacts on
agriculture (Heng et al., 2010). Thus, several technologies were developed by soil
scientists to obtain detailed information on soil surface variation caused by
erosion (Nouwakpo and Huang, 2012).
Contact techniques such as erosion pins and rillmeter have been used for
a long time to study the changes of soil surface during the erosion process (Elliot
et al., 1997; Kronvang et al., 2012). However, although the observed change in the
exposed part of the pin can be used to calculate the amount of erosion after an
erosive event, the accuracy of this technique is limited by the low spatial coverage
12
due to the small number of pins installed in the monitored area (Sirvent et al.
1997; Zhang et al., 2011). Otherwise, the rillmeter technique can obtain accurate
soil surface change data but may affect the results due to the contact between the
device and the soil surface (Elliot et al., 1997). Simple topographic instruments
(theodolites or total stations) with contact with the measured surface were also
used for a long time to study soil topography (Moser et al., 2007).
Measurement techniques of soil microtopography change that do not
involve surface contact during measurements are preferred as they do not cause
any impact on the soil and reduce the time of data acquisition (Jester and Klik,
2005). Laser scanning and digital photogrammetry have been the most used
technologies among non-contact methodologies to obtain soil surface
topographic data at high-resolution.
1.3.1 High-resolution topographic methods
High-resolution topographic data is of great importance in almost all
applications of geosciences (James and Quinton, 2014; Agüera-Vega et al., 2018;
Deng et al., 2018; Azareh et al., 2019). Thus, a wide variety of methods have
evolved to meet these surveys demands such as aerial and terrestrial laser
scanning, multibeam SONAR, RTK-DGPS and total station (Brasington, 2010;
Höfle and Rutzinger, 2011; Hohenthal et al., 2011; Castillo et al., 2012; Day et al.,
2013; James and Quinton, 2014; Vinci et al., 2015).
13
However, despite the diversity of methods available, the generation of
high-resolution DEMs requires large investments in personal training and
equipment. Thus, with the emergence of image-based methods such as digital
photogrammetry, it has drastically reduced these operating costs (Westoby et al.,
2012).
1.3.2 Digital photogrammetry
Digital photogrammetry is becoming increasingly accessible to
researchers and users, due to the development of methods that allow accurate
calibration of non-metric cameras and the reliable automation of the
photogrammetric process (Fonstad et al., 2013). Since the emergence of aerial and
digital close-range digital photogrammetry, this powerful technique has been
widely used in obtaining 3-D soil surface models (Rieke-Zapp and Nearing, 2005;
Gessesse et al., 2010; Heng et al., 2010; Nouwakpo and Huang, 2012; Stöcker et
al., 2015; Guo et al., 2016; Goetz et al., 2018).
Recent advances in digital close-range photogrammetry technologies, as
well as computer vision, have made it possible to generate high-resolution soil
topography models from consumer-grade digital cameras (Berger et al., 2010;
Heng et al., 2010; Nouwakpo and Huang, 2012; Nouwakpo et al., 2014; Guo et
al., 2016). Digital photogrammetry has the advantage of having a low cost of
equipment acquisition, with values several orders of magnitude lower than laser
scanner (Nouwakpo et al., 2014).
14
The reduction in cost and improvements in the quality of compact cameras
and single lens reflex (SLR) has popularized the access to photogrammetric
modelling and encouraged the use in several areas of the geosciences (Lane, 2000;
Chandler et al., 2002; Brasington and Smart, 2003; Marzolff and Poesen, 2009;
Bird et al., 2010). Thus, the digital close-range photogrammetry technique has
been widely used to generate DEMs with sufficient resolution for soil
microtopography studies (Babault et al., 2004; Rieke-Zapp and Nearing, 2005;
Aguilar et al., 2009; Nouwakpo and Huang, 2012; Guo et al., 2016).
Several studies have demonstrated the efficiency of the high-resolution
digital close-range photogrammetry in monitoring soil erosion, producing DEMs
with resolution of grids ranging from 1 to 15 mm (Brasington and Smart, 2003;
Abd Elbasit et al., 2009; Aguilar et al., 2009; Rieke-Zapp and Nearing, 2005; Heng
et al., 2010; Stöcker et al., 2015; Guo et al., 2016).
1.3.3 Unmanned aerial vehicles (UAV)
In order to overcome the limitations of traditional photogrammetry
techniques, the use of photographic cameras coupled in unmanned aerial
vehicles (UAVs) to acquire images of the soil surface has been objecting of study
in the last years (Bemis et al., 2014; James and Robson , 2014; Nex and
Remondino, 2014; Clapuyt et al., 2016; Di Stefano et al., 2017; O'Connor et al.,
2017; Eltner et al., 2018). UAVs have some advantages over piloted aircraft and
satellites, especially in relation to low cost, operational flexibility and better
15
spatial and temporal resolution of the images from which DEMs can be derived
(Laliberte et al., 2010; Harwin and Lucieer, 2012; Anderson and Gaston, 2013;
Hugenholtz et al., 2015; Balek and Blahůt, 2017).
UAVs require less time to acquire data when compared to other
techniques, reducing operating costs. Moreover, the resolution and accuracy of
the results obtained by UAVs cannot be achieved through satellite images
(Immerzeel, et al., 2014), being useful mainly in places where the use of other
techniques is not feasible or dangerous. The UAVs provide a convenient remote
sensing platform for studies of landslides due to their ability to acquire high-
resolution images on terrains difficult to access (Lucieer et al., 2014).
1.3.4 Structure-from-Motion (SfM)
The structure-from-motion (SfM) is based on the basic principles of
traditional photogrammetry, where the three-dimensional (3-D) structure can be
reconstructed from a series of overlapping images (Westoby et al., 2012). This
technique combines the computational vision approaches SfM (Ullman, 1979)
and multiview-stereo (MVS; Seitz et al., 2006) algorithms. However, it differs
from conventional photogrammetry as the scene geometry, orientations and
camera positions are estimated automatically, without the need to specify in
advance targets with known 3-D positions.
These parameters are computed simultaneously using the highly
redundant, iterative bundle adjustment procedure, based on the resources
16
database extracted automatically from the set of overlapping images (Snavely,
2008). This technique is suitable for sets of images that have a high degree of
overlap, with the capture of the entire 3-D structure of the scene from a wide
variety of positions (James and Robson, 2012), or according to the name, images
obtained from a moving sensor.
Based on the calculations of orientation and location of the camera,
overlapping images are used to reconstruct a 3-D sparse point cloud from the
surface. However, this result is still insufficiently detailed and very noisy for
high-quality 3-D reconstruction. In the last years, 3-D reconstruction software has
increased the quality of the models by linking the output of SfM with MVS image
matching algorithms. The MVS process effectively filters out the noise and
greatly increases the density of reconstructed points (James and Robson, 2012).
Thus, the point clouds resulting from SfM-MVS processing are of sufficient
quality to generate high-resolution DEMs but do not have geospatial or scale
information. In this way, measurements of ground control points are necessary
for later georeferencing of the models.
Studies show that the use of SfM combined with MVS can produce high-
resolution 3-D models (centimetres to millimetres), with results similar to those
generated from airborne LiDAR and terrestrial laser scanner (TLS) (Castillo et al.,
2012; James and Quinton, 2014; Eltner et al., 2015). However, there are no studies
evaluating the accuracy of eroded volumes estimates via UAV-SfM and
comparing them with independent measurements by collecting eroded soil in a
controlled environment, such as erosion plots. The validation of UAV-SfM for
17
monitoring thin surface layers of erosion opens up opportunities in the use of
this technique to elaborate strategies to control complex erosive environments,
where sheet erosion, grooves and gullies occur at the same time.
Recent studies were carried out using UAV images and SfM techniques
for soil surface mapping. The applications include monitoring of soil
microtopography change detection (Eltner et al., 2018), rill and interrill erosion
(Eltner et al., 2015), gully erosion (Gómez-Gutiérrez et al., 2014; Stöcker et al.,
2015; Glendell et al., 2017), wind erosion (Pagán et al., 2019) and landslides
(Lucieer et al., 2014; Turner et al., 2015). However, there is a gap of work using
UAV-SfM to evaluate the behaviour of laminar erosion in the field, as well as the
relative contributions of different types of erosion in the gully development.
In this thesis, the SfM-MVS technique will be treated only as SfM,
following the trend of the recent works in the area of geosciences (Hänsel et al.,
2016; Cook, 2017; Mlambo et al., 2017; Mosbrucker et al., 2017; Agüera-Vega et
al., 2018; Zimmer et al., 2018).
1.4 Thesis aims and objectives
This project aims to investigate the efficiency of aerial and terrestrial
photogrammetry in the detailed study of water erosion, as well as to address
some gaps in our understanding of the evolution of erosive processes in its
different forms (sheet, rill and gully erosion) of occurrence. In doing so, we will
evaluate the factors that influence the development of erosion in micro and
18
macro scales, with experiments in the laboratory and in the field. The hypothesis
is that the use of digital photogrammetry will allow soil scientists to accurately
quantify laminar erosion, making it possible to replace traditional methods, such
as erosion plots, with high operating costs. In addition, high-resolution
photogrammetry will make it possible to study the dynamics of evolution of the
gullies in space and time, allowing to evaluate the relative contribution of sheet
erosion and rills to the growth of the gullies complex.
This thesis sets out to develop an understanding at both a methodological
level and in terms of wider soil erosion processes understandings. From a tightly
controlled study to field observations, it will study the accuracy and application
of ground and air-based photogrammetry for soil erosion assessment at a range
of scales. The thesis consists of three experimental chapters designed to address
the following objectives:
(i) To evaluate the influence of gradient change and runoff volumes
on rill erosion process, using digital close-range photogrammetry in a laboratory
soil flume. In addition, to estimate morphological rill parameters for a better
understanding of the rill erosion behaviour under different treatments. (Chapter
1)
(ii) To evaluate the efficiency of SfM based on UAV images in obtaining
accurate measurements of soil loss in areas of sheet erosion, under natural
rainfall, where channelized erosion is not the principal mechanism. The
19
measurements acquired from SfM were compared to the sediments collected in
each soil erosion plots. (Chapter 2)
(iii) A case study of a gully in Brazil using SfM to provide a detailed
understanding of the dynamics of the sediment movements in the gully system.
Thus, to evaluate the relative contribution of sheet and rill erosion, as well as
gully sidewalls to sediment generation. In addition, to quantify the total volume
of sediments stored and lost from the gully system over time. (Chapter 3)
1.5 References
Abd Elbasit, M.A., Anyoji, H., Yasuda, H., Yamamoto, S., 2009. Potential of low cost
close-range photogrammetry system in soil microtopography quantification. Hydrol.
Process. 23, 1408–1417.
Agüera-Vega, F., Carvajal-Ramírez, F., Martínez-Carricondo, P., López, J.S-H., Mesas-
Carrascosa, F.J., García-Ferrer, A., Pérez-Porras, F.J., 2018. Reconstruction of extreme
topography from UAV structure from motion photogrammetry. Measurement 121, 127–
138.
Aguilar, M.A., Aguilar, F.J., Negreiros, J., 2009. Off-the-shelf laser scanning and close-
range digital photogrammetry for measuring agricultural soils microrelief. Biosyst. Eng.
103, 504–517.
Aksoy, H., Unal, N.E., Cokgor, S., Gedikli, A., Yoon, J., Koca, K., Inci, S.B., Eris, E., Pak,
G., 2013. Laboratory experiments of sediment transport from bare soil with a rill. Hydrol.
Sci. J. 58, 1505–1518.
Alexandratos, N., Bruinsma, J., 2012. World agriculture towards 2030/2050: the 2012
revision. ESA Working paper No. 12-03. Rome, FAO.
Allen, P.M., Arnold, J.G., Auguste, L., White, J., Dunbar, J., 2018. Application of a simple
headcut advance model for gullies. Earth Surf. Process. Landf. 43, 202–217.
Anderson, K., Gaston, K.J., 2013. Lightweight unmanned aerial vehicles will
revolutionize spatial ecology. Front. Ecol. Environ. 11, 138–146.
Azareh, A., Rahmati, O., Rafiei-Sardoo, E., Sankey, J.B., Lee, S., Shahabi, H., Ahmad, B.B.,
2019. Modelling gully-erosion susceptibility in a semi-arid region, Iran: Investigation of
20
applicability of certainty factor and maximum entropy models. Sci. Total Environ. 655,
684–696.
Babault, J., Bonnet, S., Crave, A., Van Den Driessche, J., 2004. Influence of piedmont
sedimentation on erosion dynamics of an uplifting landscape: an experimental
approach. Geology 33, 301–304.
Bagnold, R.A., 1966. An approach to the sediment transport problem from general
physics. USGS Prof. Pap. 442-I. U.S. Gov. Print. Office, Washington, DC.
Balek, J., Blahůt, J., 2017. A critical evaluation of the use of an inexpensive camera
mounted on a recreational unmanned aerial vehicle as a tool for landslide research,
Landslides 14, 1217–1224.
Bastola, S., Dialynas, Y.G., Bras, R.L., Noto, L.V., Istanbulluoglu, E., 2018. The role of
vegetation on gully erosion stabilization at a severely degraded landscape: a case study
from Calhoun Experimental Critical Zone Observatory. Geomorphology 308, 25–39.
Bazoffi, P., 2015. Measurement of rill erosion through a new UAV-GIS methodology.
Ital. J. Agron. 10 (s1), 708.
Bemis, S.P., Micklethwaite, S., Turner, D., James, M.R., Akciz, S., Thiele, S.T., Ali
Bangash, H., 2014. Ground-based and UAV-Based photogrammetry: a multi-scale, high-
resolution mapping tool for structural geology and paleoseismology. J. Struct. Geol. 69,
P163–P178.
Ben Slimane, A., Raclot, D., Rebai, H., Le Bissonnais, Y., Planchon, O., Bouksila F., 2018.
Combining field monitoring and aerial imagery to evaluate the role of gully erosion in a
Mediterranean catchment (Tunisia). Catena 170, 73–83.
Berger, C., Schulze, M., Rieke-Zapp, D., Schlunegger, F., 2010. Rill development and soil
erosion: a laboratory study of slope and rainfall intensity. Earth Surf. Process. Landf. 35,
1456–1467.
Bird, S., Hogan, D., Schwab, J., 2010. Photogrammetric monitoring of small streams
under a riparian forest canopy. Earth Surf. Process. Landf. 35, 952–970.
Blanco, H., Lal, R., 2010. Principles of soil conservation and management. Springer,
Heidelberg, Germany, 620 pp.
Bollati, I., Vergari, F., DelMonte,M., Pelfini,M., 2016. Multitemporal
dendrogeomorphological analysis of slope instability in upper Orcia Valley (Southern
Tuscany, Italy). Geogr. Fis. Din. Quat. 39, 105–120.
Brasington, J., 2010. From grain to floodplain: hyperscale models of braided rivers. J.
Hydraul. Res. 48, 52–53.
Brasington, J., Smart, R.M.A., 2003. Close range digital photogrammetric analysis of
experimental drainage basin evolution. Earth Surf. Process. Landf. 28, 231–247.
21
Brunton, D.A., Bryan, R.B., 2000. Rill network development and sediment budgets. Earth
Surf. Process. Landf. 25, 783–800.
Bryan, R.B., 1990. Knickpoint evolution in rillwash. Catena Supplement 17, 111–132.
Bryan, R.B., Rockwell, D.L., 1998. Water table control on rill initiation and implications
for erosional response. Geomorphology 23, 151–169.
Castillo, C., Pérez, R., James, M.R., Quinton, J.N., Taguas, E.V., Gómez, J.A., 2012.
Comparing the accuracy of several field methods for measuring gully erosion. Soil Sci.
Soc. Am. J. 76, 1319–1332.
Cerdan, O., Govers, G., Le Bissonnais, Y., Van Oost, K., Poesen, J., Saby, N., Gobin, A.,
Vacca, A., Quinton, J., Auerswald, K., Klik, A., Kwaad, F., Raclot, D., Ionita, I., Rejman,
J., Rousseva, S., Muxart, T., Roxo, M., Dostal, T., 2010. Rates and spatial variations of soil
erosion in Europe: A study based on erosion plot data. Geomorphology 122, 167–177.
Cerdan, O., Le Bissonnais, Y., Couturier, A., Saby, N., 2002. Modelling interrill erosion
in small cultivated catchments. Hydrol. Process. 16, 3215–3226.
Chandler, J., Ashmore, P., Paola, C., Gooch, M., Varkaris, F., 2002. Monitoring river-
channel change using terrestrial oblique digital imagery and automated digital
photogrammetry. Ann. Assoc. Am. Geogr. 92, 631–644.
Chaplot, V., 2013. Impact of terrain attributes, parent material and soil types on gully
erosion. Geomorphology 186, 1–11.
Clapuyt, F., Vanacker, V., Van Oost, K., 2016. Reproducibility of UAV-based earth
topography reconstructions based on Structure-from-Motion algorithms.
Geomorphology 260, 4–15.
Conoscenti, C., Angileri, S., Cappadonia, C., Rotigliano, E., Agnesi, V., Märker, M., 2014.
Gully erosion susceptibility assessment by means of GIS-based logistic regression: a case
of Sicily (Italy). Geomorphology 204, 399–411.
Cook, K.L., 2017. An evaluation of the effectiveness of low-cost UAVs and structure from
motion for geomorphic change detection. Geomorphology 278, 195–208.
Day, S.S., Gran, K.B., Belmont, P., Wawrzyniec, T., 2013. Measuring bluff erosion part 2:
pairing aerial photographs and terrestrial laser scanning to create a watershed scale
sediment budget. Earth Surf. Process. Landf. 38, 1068–1082.
Defersha, M.B., Melesse, A.M., 2012. Effect of rainfall intensity, slope and antecedent
moisture content on sediment concentration and sediment enrichment ratio. Catena 90,
47–52.
Deng, L., Mao, Z., Li, X., Hu, Z., Duan, F., Yan, Y., 2018. UAV-based multispectral remote
sensing for precision agriculture: A comparison between different cameras. ISPRS-J.
Photogramm. Remote Sens. 146, 124–136.
22
Descroix L., González Barrios, J.L., Viramontes, D., Poulenard, J., Anaya, E., Esteves, M.,
Estrada, J., 2008. Gully and sheet erosion on subtropical mountain slopes: Their
respective roles and the scale effect. Catena 72, 325–339.
Devente, J., Poesen, J., 2005. Predicting soil erosion and sediment yield at the basin scale:
scale issues and semi-quantitative models. Earth Sci. Rev. 71, 95–125.
Di Stefano, C., Ferro, V., Palmeri, V., Pampalone, V., 2017. Measuring rill erosion using
structure from motion: A plot experiment. Catena 156, 383–392.
Di Stefano, C., Ferro, V., Pampalone, V., Sanzone, F., 2013. Field investigation of rill and
ephemeral gully erosion in the Sparacia experimental area, South Italy. Catena 101, 226–
234.
Dlamini, P., Orchard, C., Jewitt, G., Lorentz, S., Titshall, L., Chaplot, V., 2011. Controlling
factors of sheet erosion under degraded grasslands in the sloping lands of KwaZulu-
Natal, South Africa. Agric. Water Manage. 98, 1711–1718.
Elliot, W., Laflen, J., Thomas, A., Kohl, K., 1997. Photogrammetric and rillmeter
techniques for hydraulic measurement in soil erosion studies. Trans. ASAE 40, 157–165.
Ellison, W.D., 1947. Soil erosion studies: I. Agric. Eng. 28, 145–146.
Eltner, A., Baumgart, P., Maas, H.G., Faust, D., 2015. Multi-temporal UAV data for
automatic measurement of rill and interrill erosion on loess soil. Earth Surf. Process.
Landf. 40, 741–755.
Eltner, A., Maas, H-G., Faust, D., 2018. Soil micro-topography change detection at
hillslopes in fragile Mediterranean landscapes. Geoderma 313, 217–232.
Eswaran, H., Lal, R., Reich, P.F., 2001. Land degradation: An overview. In: E.M. Bridges,
F.W.T. Penning de Vries, L.R. Oldeman, S. Sombatpanit, S.J. Scherr (Eds.), Response to
land degradation. Science Publishers, Enfield, pp. 20-35.
Fonstad, M.A., Dietrich, J.T., Courville, B.C., Jensen, J.L., Carbonneau, P.B., 2013.
Topographic structure from motion: a new development in photogrammetric
measurement. Earth Surf. Process. Landf. 38, 421–430.
Fournier, F., 1960. Climate et érosion. Presses Universitaires de France, Paris. 201 pp.
García-Ruiz, J.M., Beguería, S., Nadal-Romero, E., Gonzalez-Hidalgo, J.C., Lana-Renault,
N., Sansjuan, Y., 2015. A meta-analysis of soil erosion rates across the world.
Geomorphology 239, 160–173.
Gessesse, G.D., Fuchs, H., Mansberger, R., Klik, A., Rieke-Zapp, D.H., 2010. Assessment
of erosion, deposition and rill development on irregular soil surfaces using close range
digital photogrammetry. Photogramm. Rec. 25, 299–318.
Glendell, M., Brazier, R.E., 2014. Accelerated export of sediment and carbon from a land-
scape under intensive agriculture. Sci. Total Environ. 476–477, 643–656.
23
Glendell, M., McShane, G., Farrow, L., James, M., Quinton, J., Anderson, K., Evans, M.,
Benaud, P., Rawlins, B., Morgan, D., Jones, L., Kirkham, M., DeBell, L., Quine, T., Lark,
M., Rickson, J., Brazier, R., 2017. Testing the utility of structure from motion
photogrammetry reconstructions using small unmanned aerial vehicles and ground
photography to estimate the extent of upland soil erosion. Earth Surf. Process. Landf. 42,
1860–1871.
Goetz, J., Brenning, A., Marcer, M., Bodin, X., 2018. Modeling the precision of structure-
from-motion multi-view stereo digital elevation models from repeated close-range aerial
surveys. Remote Sens. Environ. 210, 208–216.
Gómez-Gutiérrez A, Schnabel S, Berenguer-Sempere F, Lavado- Contador F, Rubio-
Delgado J. 2014. Using 3D photo-reconstruction methods to estimate gully headcut
erosion. Catena 120, 91–101.
Govers G, Giménez R, Van Oost K. 2007. Rill erosion: exploring the relationship between
experiments, modelling and field observations. Earth-Sci. Rev. 84, 87–102.
Guo, Q., Hao, Y., Liu, B., 2015. Rates of soil erosion in China: A study based on runoff
plot data. Catena 2015, 68–76.
Guo, Z.L., Ma, M.J., Cai, C.F., Wu, Y.W., 2017. Combined effects of simulated rainfall
and overland flow on sediment and solute transport in hillslope erosion. J. Soils
Sediments 4, 1–13.
Guo, M., Shi, H., Zhao, J., Liu, P., Welbourne, D., Lin, Q., 2016. Digital close range
photogrammetry for the study of rill development at flume scale. Catena 143, 265–274.
Hancock, G.R., Loughran, R.J., Evans, K.G., Balog, R.M., 2008. Estimation of soil erosion
using field and modelling approaches in an undisturbed Arnhem Land catchment,
Northern Territory, Australia. Geogr. Res. 46, 333–349.
Hänsel, P., Schindewolf, M., Eltner, A., Kaiser, A., Schmidt, J., 2016. Feasibility of high-
resolution soil erosion measurements by means of rainfall simulations and SfM
photogrammetry. Hydrology 3 (38).
Hao, H., Wang, J., Guo, Z., Hua, L., 2019. Water erosion processes and dynamic changes
of sediment size distribution under the combined effects of rainfall and overland flow.
Catena, 173, 494–504.
Harvey, A.M., 1992. Process interactions, temporal scales and the development of
hillslope gully systems: Howgill Fells, northwest England. Geomorphology 5, 323–344.
Harwin, S., Lucieer, A., 2012. Assessing the accuracy of georeferenced point clouds
produced via multi-view stereopsis from unmanned aerial vehicle (UAV) imagery.
Remote Sens. 4, 1573–1599.
He, J.J., Sun, L.Y., Gong, H.L., Cai, Q.G., Jia, L.J., 2016. The characteristics of rill
development and their effects on runoff and sediment yield under different slope
gradients. J. Mt. Sci. 13, 397–404.
24
Heng, B.C.P., Chandler, J.H., Armstrong, A., 2010. Applying close-range digital
photogrammetry and soil erosion studies. The Photogrammetric Record 25, 240–265.
Heras, M.M., Espigares, T., Merino-Martín, L., Nicolau, J.M., 2011. Water-related
ecological impacts of rill erosion processes in Mediterranean-dry reclaimed slopes.
Catena 84, 114–124.
Höfle, B., Rutzinger, M., 2011. Topographic airborne LiDAR in geomorphology: a
technological perspective. Z. Geomorphol. 55, 1–29.
Hohenthal, J., Alho, P., Hyyppa, J., Hyyppa, H., 2011. Laser scanning applications in
fluvial studies. Prog. Phys. Geogr. 35, 782–809.
Horton, R.E., 1945. Erosional development of streams and their drainage basins;
hydrophysical approach to quantitative morphology. Geol. Soc. Am. Bull. 56, 275–370.
Hosseinalizadeh, M., Kariminejad, N., Chen, W., Pourghasemi, H.R., Alinejad, M.,
Behbahani, A.M., Tiefenbacher, J.P., 2019. Spatial modelling of gully headcuts using
UAV data and four best-first decision classifier ensembles (BFTree, Bag-BFTree, RS-
BFTree, and RF-BFTree). Geomorphology 329, 184–193.
Hugenholtz, C.H., Walker, J., Brown, O., Myshak, S., 2015. Earthwork volumetrics with
an unmanned aerial vehicle and softcopy photogrammetry, J. Surv. Eng. 141, 06014003
Ibáñez, J., Lavado Contador, J.F., Schnabel, S., Martínez, V.J., 2016. Evaluating the
influence of physical, economic and managerial factors on sheet erosion in rangelands
of SW Spain by performing a sensitivity analysis on an integrated dynamic model. Sci.
Total Environ. 544, 439–449.
Immerzeel, W.W., Kraaijenbrink, P.D.A., Shea, J.M., Shrestha, A.B., Pellicciotti, F.,
Bierkens, M.F.P., de Jong, S.M., 2014. High-resolution monitoring of Himalayan glacier
dynamics using unmanned aerial vehicles. Remote Sens. Environ. 150, 93–103.
Ionita, I., Fullen, M.A., Zgłobicki, W., Poesen, J., 2015. Gully erosion as a natural and
human-induced hazard. Nat. Hazards 79, 1–5.
James, M.R., Quinton, J.N., 2014. Ultra-rapid topographic surveying for complex
environments: the hand-held mobile laser scanner (HMLS). Earth Surf. Process. Landf.
39, 138–142.
James, M.R., Robson, S., 2012. Straightforward reconstruction of 3D surfaces and
topography with a camera: Accuracy and geoscience application. J. Geophys. Res. 117,
F03017.
James, M.R., Robson, S., 2014. Mitigating systematic error in topographic models derived
from UAV and ground-based image networks. Earth Surf. Process. Landf. 39, 1413–1420.
Jester, W., Klik, A., 2005. Soil surface roughness measurement – methods, applicability,
and surface representation. Catena 64, 174–192.
25
Jiang, F., Zhan, Z., Chen, J., Lin, J., Wang, M.K., Ge, H., 2018. Rill erosion processes on a
steep colluvial deposit slope under heavy rainfall in flume experiments with artificial
rain. Catena 169, 46–58.
Kaiser, A., Neugirg, F., Rock, G., Müller, C., Haas, F., Ries, J., Schmidt, J., 2014. Small-
scale surface reconstruction and volume calculation of soil erosion in complex Moroccan
gully morphology using structure from motion. Remote Sens. 6, 7050–7080.
Kinnell, P.I.A., 2006. Simulations demonstrating interaction between coarse and fine
sediment loads in rain-impacted flow. Earth Surf. Process. Landf. 31, 355–367.
Kinnell, P.I.A., 2013. Modeling of the effect of flood depth on sediment discharged by
rain-impacted flows from sheet and interrill erosion areas: a review. Hydrol. Process. 27,
2567–2578.
Kirkby, M.J., Bracken, L.J., 2009. Gully processes and gully dynamics. Earth Surf.
Process. Landf. 34, 1841–1851.
Kronvang, B., Audet, J., Baattrup-Pedersen, A., Jensen, H.S., Larsen, S.E., 2012.
Phosphorus load to surface water from bank erosion in a Danish lowland river basin. J.
Environ. Qual. 41, 304–313.
Lal, R., 2001. Soil degradation by erosion. Land Degrad. Dev. 12, 519–529.
Lal, R., 2003. Soil erosion and the global carbon budget. Environ. Int. 29, 437–450.
Lal, R. 2013. Climate-strategic agriculture and the water-soil-waste nexus. J. Plant Nutr.
Soil Sci. 176, 479-493.
Laliberte, A., Herrick, J., Rango, A., Winters, C., 2010. Acquisition, orthorectification, and
object-based classification of unmanned aerial vehicle (UAV) imagery for rangeland
monitoring. Photogramm. Eng. Remote. Sens. 76, 661–672.
Lane, S.N., 2000. The measurement of river channel morphology using digital
photogrammetry. Photogramm. Rec. 16, 937–957.
Le Bissonnais, Y., 1996. Aggregate stability and assessment of soil crustibility: I. Theory
and methodology. Eur. J. Soil Sci. 47, 425–437.
Le Bissonnais, Y., Benkhadra, H., Chaplot, V., Fox, D., King, D., Daroussin, J., 1998.
Crusting, runoff and sheet erosion on silty loamy soils at various scales and upscaling
from m2 to small catchments. Soil Tillage Res. 46, 69–80.
Le Bissonnais, Y., Cerdan, O., Lecomte, V., Benkhadra, H., Souchère, V., Martin, P., 2005.
Variability of soil surface characteristics influencing runoff and interrill erosion. Catena
62, 111–124.
Lu, X., Li, Y., Washington-Allen, R.A., Li, Y., Li, H., Hu, Q., 2017. The effect of grid size
on the quantification of erosion, deposition, and rill network. Int. Soil Water Cons. Res.
5, 241–251.
26
Lucieer, A., Jong, S.M.d., Turner, D., 2014. Mapping landslide displacements using
Structure from Motion (SfM) and image correlation of multi-temporal UAV
photography. Prog. Phys. Geogr. 38, 97–116.
Luffman, I.E., Nandi, A., Spiegel, T., 2015. Gully morphology, hillslope erosion, and
precipitation characteristics in the Appalachian Valley and Ridge province, southeastern
USA. Catena 133, 221–232.
Mahmoodabadi, M., Sajjadi, S.A., 2016. Effects of rain intensity, slope gradient and
particle size distribution on the relative contributions of splash and wash loads to rain-
induced erosion. Geomorphology 253, 159–167.
Maïga-Yaleu, S.B., Chivenge, P., Yacoub, H., Guiguemde, I., Karambiri, H., Ribolzi, O.,
Bary, A., Chaplot, V., 2015. Impact of sheet erosion mechanisms on organic carbon losses
from crusted soils in the Sahel. Catena 126, 60–67.
Makanzu Imwangana, F., Dewitte, O., Ntombi, M., Moeyersons, J., 2014. Topographic
and road control of mega-gullies in Kinshasa (DR Congo). Geomorphology 217, 131–139.
Mancilla, G.A., Chen, S., Mccool, D.K., 2005. Rill density prediction and flow velocity
distributions on agricultural areas in the Pacific Northwest. Soil Tillage Res. 84, 54–66.
Marzolff, I., Poesen, J., 2009. The potential of 3D gully monitoring with GIS using high
resolution aerial photography and a digital photogrammetry system. Geomorphology
111, 48–60.
Mitas, L., Mitasova, H., 1998. Distributed soil erosion simulation for effective erosion
prevention. Water Resour. Res. 34, 505–516.
Mlambo, R., Woodhouse, I., Gerard, F., Anderson, K., 2017. Structure from motion (SfM)
Photogrammetry with drone data: a low cost method for monitoring greenhouse gas
emissions from forests in developing countries. Forests 8, 68.
Morgan, R.P.C., 1977. Soil Erosion in the United Kingdom: Field studies in the Silsoe
area, 1973–1975. Occasional Paper. 4 Nat. Coll. Agr. Eng. Silsoe, UK.
Morgan, J.A., Brogan, D.J., Nelson, P.A., 2017. Application of Structure-from-Motion
photogrammetry in laboratory flume. Geomorphology 276, 125–143.
Mosbrucker, A.R., Major, J.J., Spicer, K.R., Pitlick, J., 2017. Camera system considerations
for geomorphic applications of SfM photogrammetry. Earth Surf. Process. Landf. 42,
969–986.
Moser, K., Ahn, C., Noe, G., 2007. Characterization of microtopography and its influence
on vegetation patterns in created wetlands. Wetlands, 27, 1081–1097.
Müller-Nedebock, D., Chivenge, P., Chaplot, V., 2016. Selective organic carbon losses
from soils by sheet erosion and main controls. Earth Surf. Process. Landf. 41, 1399–1408.
Nachtergaele, J., Poesen, J., 2002. Spatial and temporal variations in resistance of loess-
derived soils to ephemeral gully erosion. Eur. J. Soil Sci. 53, 449–463.
27
Neugirg, F., Stark, M., Kaiser, A., Vlacilova, M., Della Seta, M., Vergari, F., Schmidt, J.,
Becht, M., Haas, F., 2016. Erosion processes in calanchi in the upper Orcia Valley,
southern Tuscany, Italy based on multitemporal high-resolution terrestrial LiDAR and
UAV surveys. Geomorphology 269, 8–22.
Nex, F., Remondino, F., 2014. UAV for 3D mapping applications: a review. Appl.
Geomatics 6, 1–15.
Nouwakpo, S.K., Huang, C.-H., 2012. A simplified close-range photogrammetric
technique for soil erosion assessment. Soil Sci. Soc. Am. J. 76, 70–84.
Nouwakpo, S.K., James, M.R., Weltz, M.A., Huang, C.-H., Chagas, I., Lima, L., 2014.
Evaluation of structure from motion for soil microtopography measurement.
Photogramm. Rec. 29, 297–316.
Nouwakpo, S., Weltz, M., McGwire, K., 2015. Assessing the performance of structure-
from-motion photogrammetry and terrestrial lidar for reconstructing soil surface
microtopography of naturally vegetated plots. Earth Surf. Process. Landf. 41, 308–322.
O’Connor, J., Smith, M.J., James, M.R., 2017. Cameras and settings for aerial surveys in
the geosciences: Optimising image data. Prog. Phys. Geogr. 41, 325–344.
Pagán, J.I., Bañon, L., López, I., Bañon, C., Aragonés, L., 2019. Monitoring the dune-beach
system of Guardamar del Segura (Spain)using UAV, SfM and GIS techniques. Sci. Total
Environ. 687, 1034–1045.
Paul Obade, V., Lal, R., 2016. Towards a standard technique for soil quality assessment.
Geoderma 265, 96-102.
Perroy, R.L., Bookhagen, B., Asner, G.P., Chadwick, O.A., 2010. Comparison of gully
erosion estimates using airborne and ground-based LIDAR on Santa Cruz Island,
California. Geomorphology 118, 288–300.
Pimentel, D., 2006. Soil erosion: a food and environmental threat. Environ. Dev. Sustain.
8, 119-137.
Poesen, J., Nachtergaele, J., Verstraeten, G., Valentin, C., 2003. Gully erosion and
environmental change: importance and research needs. Catena 50, 91–133.
Poesen, J., Vanwalleghem, T., Deckers, J., 2018. Gullies and closed depressions in the
Loess Belt: scars of human–environment interactions. Landscapes and Landforms of
Belgium and Luxembourg. Springer, Cham, pp. 253–267.
Polyakov, V.O., Nearing, M.A., 2003. Sediment transport in rill flow under deposition
and detachment conditions. Catena 51, 33–43.
Rauws, G., 1987. The initiation of fills on plane beds of non-cohesive sediments. Catena
Supplement 8, 107–118.
Renard, K.G., Foster, G.R., Weesies, G.A., McCool, D.K., Yoder, D.C., 1997. Predicting
soil erosion by water: a guide to conservation planning with the revised universal soil
28
loss equation (RUSLE). Agriculture Handbook 703US Department of Agriculture,
Washington, DC.
Rieke-Zapp, D.H., Nearing, M.A., 2005. Digital close range photogrammetry for
measurement of soil erosion. Photogramm. Rec. 20, 69–87.
Rodrigo Comino, J., Brings, C., Lassu, T., Iserloh, T., Senciales, J.M., Martínez Murillo,
J.F., Ruiz Sinoga, J.D., Seeger, M., Ries, J.B., 2015. Rainfall and human activity impacts
on soil losses and rill erosion in vineyards (Ruwer Valley, Germany). Solid Earth 6, 823–
837
Rickson, R.J., 2014. Can control of soil erosion mitigate water pollution by sediments?
Sci. Total Environ. 468–469, 1187–1197.
Sankey, J.B., Draut, A.E., 2014. Gully annealing by aeolian sediment: field and remote-
sensing investigation of aeolian–hillslope–fluvial interactions, Colorado River corridor,
Arizona, USA. Geomorphology 220, 68–80.
Seitz, S.M., Curless, B., Diebel, J., Scharstein, D., Szeliski, R., 2006. A comparison and
evaluation of multi-view stereo reconstruction algorithms. Computer Vision and Pattern
Recognition, 2006 IEEE Computer Society Conference on, pp. 519–528.
Shi, Z.H., Yan, F.L., Li, L., Li, Z.X., Cai, C.F., 2010. Interrill erosion from disturbed and
undisturbed samples in relation to topsoil aggregate stability in red soils from
subtropical China. Catena 81, 240–248.
Sidorchuk, A., Marker, M., Moretti, S., Rodolfi, G., 2003. Gully erosion modelling and
landscape response in the Mbuluzi River catchment of Swaziland. Catena 50, 507–525.
Sirjani, E., Mahoodabadi, M., 2014. Effects of sheet flow rate and slope gradient on
sediment load. Arab. J. Geosci. 7, 203–210.
Sirvent, J., Desir, G., Gutierrez, M., Sancho, C., Benito, G., 1997. Erosion rates in badland
areas recorded by collectors, erosion pins and profilometer techniques (Ebro Basin, NE-
Spain). Geomorphology 18, 61–75.
Snavely, N., Seitz, S., Szeliski, R., 2008. Modeling the world from internet photo
collections. Int. J. Comput. Vis. 80, 189–210.
Sofia, G., Di Stefano, C., Ferro, V., Tarolli, P., 2017. Morphological similarity of channels:
from linear erosional features (rill, gully) to alpine rivers. Land Degrad. Dev. 28, 1717–
1728.
Stöcker, C., Eltner, A., Karrasch, P., 2015. Measuring gullies by synergetic application of
UAV and close range photogrammetry – A case study from Andalusia, Spain. Catena
132,1–11.
Thorburn, P.J., Wilkinson, S.N., 2013. Conceptual frameworks for estimating the water
quality benefits of improved agricultural management practices in large catchments.
Agric. Ecosyst. Environ. 180, 192–209.
29
Turner, D., Lucieer, A., de Jong, S., 2015. Time Series Analysis of Landslide Dynamics
Using an Unmanned Aerial Vehicle (UAV), Remote Sens. 7, 1736–1757.
Ullman, S., 1979. The interpretation of visual motion. MIT Press Series in Artificial Intelli-
gence. MIT Press, Cambridge, Massachusetts, and London, England.
Valentin, C., Poesen, J., Yong, L., 2005. Gully erosion: impacts, factors and control.
Catena 63, 132–153.
Vandekerckhove, L., Poesen, J., Wijdenes, D.O., de Figueiredo, T., 1998. Topographical
thresholds for ephemeral gully initiation in intensively cultivated areas of the
Mediterranean. Catena 33, 271–292.
Vanwalleghem, T., Poesen, J., Van Den Eeckhaut, M., Nachtergaele, J., Deckers, J., 2005.
Reconstructing rainfall and land-use conditions leading to the development of old
gullies. The Holocene 15, 378–386.
Vinci, A., Brigante, R., Todisco, F., Mannocchi, F., Radicioni, F., 2015. Measuring rill
erosion by laser scanning. Catena 124, 97–108.
Wang, B., Steiner, J., Zheng, F., Gowda, P., 2017. Impact of rainfall pattern on interrill
erosion process. Earth Surf. Process. Landf. 42, 1833–1846.
Westoby, M.J., Brasington, J., Glasser, N.F., Hambrey, M.J., Reynolds, J.M., 2012.
‘Structure-from-Motion’ photogrammetry: a low-cost, effective tool for geoscience
applications. Geomorphology 179, 300–314.
Wirtz, S., Seeger, M., Ries, J.B., 2010. The rill experiment as a method to approach a
quantification of rill erosion process activity. Z. Geomorphol. 54, 47–64.
Wirtz, S., Seeger, M., Ries, J.B., 2012. Field experiments for understanding and
quantification of rill erosion processes. Catena 91, 21–34.
Wirtz, S., Seeger, M., Zell, A., Wagner, C., Wagner, J.F., Ries, J.B., 2013. Applicability of
different hydraulic parameters to describe soil detachment in eroding rills. PLoS One 8
(5), e64861.
World Health Organization, UNICEF., 2015. Progress on Sanitation and Drinking Water:
2015 Update and Millennium Development Goals Assessment. World Health
Organization, UNICEF, Division of Communication, 3 United Nations Plaza, New York
10017, USA.
Wu, X., Wei, Y., Wang, J., Xia, J., Cai, C., Wei, Z., 2018. Effects of soil type and rainfall
intensity on sheet erosion processes and sediment characteristics along the climatic
gradient in central-south China. Sci. Total Environ. 621, 54–66.
Xu, M., Li, Q., Wilson, G., 2016. Degradation of soil physicochemical quality by
ephemeral gully erosion on sloping cropland of the hilly Loess Plateau, China. Soil
Tillage Res. 155, 9–18.
30
Zabihi, M., Mirchooli, F., Motevalli, A., Darvishan, A.K., Pourghasemi, H.R., Zakeri,
M.A., Sadighi, F., 2018. Spatial modelling of gully erosion in Mazandaran Province,
northern Iran. Catena 161, 1–13.
Zhang, J., Yang, M., Deng, X., Liu, Z., Zhang, F., 2019. The effects of tillage on sheet
erosion on sloping fields in the wind-water erosion crisscross region of the Chinese Loess
Plateau. Soil Tillage Res. 187, 235–245.
Zhang, X.C., Wang, Z.L., 2017. Interrill soil erosion processes on steep slopes. J. Hydrol.
548, 652–664.Zhang,
Zhao, J., Yang, Z., Govers, G., 2019. Soil and water conservation measures reduce soil
and water losses in China but not down to background levels: Evidence from erosion
plot data. Geoderma 337, 729–741.
Zhang, J., Zheng, F., Wen, L., 2011. Methodology of dynamic monitoring gully erosion
process using 3D laser scanning technology. Bull. Soil Water Conserv. 31, 89–94.
Zheng, F.L., Tang, K.L., 1997. Rill erosion process of steep slope land of the Loess Plateau.
International Journal of Sediment Research 12, 52–59.
Zimmer, B., Liutkus-Pierce, C., Marshall, S.T., Hatala, KG., Metallo, A., Rossi, V., 2018.
Using differential structure-from-motion photogrammetry to quantify erosion at the
Engare Sero footprint site, Tanzania. Quat. Sci. Rev. 198, 226–241.
31
2. Rill erosion and morphology development in response to changing
discharge and slope profiles
2.1 Introduction
Rill erosion is the removal of soil by concentrating water flow into a
narrow channel and, hence, increasing the sediment transport capacity of the
flow (Chen et al., 2016). It represents an intermediate process between sheet and
gully erosion (Jackson, 1997). The presence of rills in runoff plots can increase
sediment yield by 40 times (Morgan, 1977). It can lead to large soil losses, water
pollution, and damage to drainage networks (Morgan, 2005; Bewket and Sterk,
2003; Poesen et al., 2003). In agricultural lands, rill erosion can be easy to observe,
but it is difficult to measure due to its stochastic nature (He et al., 2016). During
an erosive event, the soil surface is in continuous transformation, which can
result in substantial topography variation at scales of few millimetres to as much
as a metre, depending on the volume of transported soil (Gessesse et al., 2010;
Heng et al., 2010).
Several studies have focused on rill erosion processes (Favis-Mortlock et
al., 2000; Wirtz et al., 2012), but research on rill network development remains
limited. Most study only the main rills, ignoring the important role of the
secondary channels on the rill network development (Mancilla et al., 2005; Shen
et al., 2015).
32
The amount of soil loss in sloping croplands and rangelands is affected
significantly by the rill morphology since water flow within rills has the capacity
to transport large volumes of sediment (Favis-Mortlock et al., 2000; Gatto, 2000).
Considering that rills change morphologically in time and space, it is necessary
to consider temporal and spatial variations in their development (Lei and
Nearing, 1998; Shen et al., 2015).
As rills are micro-relief channels that are generally less than 0·3 m deep
and wide (Nearing et al., 1997), a detailed spatial study requires high-resolution
digital elevation models (DEMs). Digital close-range photogrammetry has been
widely used within soil erosion studies to generate high-resolution DEMs (Abd
Elbasit et al., 2009; Aguilar et al., 2009; Guo et al., 2016; Nouwakpo and Huang,
2012; Rieke-Zapp and Nearing, 2005). This method is based on images of the soil
surface taken from multiple positions at relatively low height (Abd Elbasit et al.,
2009), enabling 3D analysis of the rill erosion development at millimetre
resolution, with instantaneous data acquisition using high-resolution consumer-
grade cameras.
However, the presence of vegetation in the study area affects the 3D
reconstruction of the soil surface by aerial images. In these cases, there are
mathematical algorithms that filter the vegetation present in the 3D point cloud,
for example CANUPO (Brodu and Lague, 2012) and CSF (Zhang et al., 2016),
allowing to estimate the soil surface with compromised accuracy. Thus, soil
surface DEMs obtained by digital close-range photogrammetry enable the
33
volumetric change detection and the assessment of microtopography change
over the time (Eltner et al., 2018).
Since conducting experiments on rill erosion processes and
microtopography change are challenging in field environments, studies that
simulate surface runoff in laboratory erosion plots are necessary to understand
the basic processes and mechanisms of rill formation, rill density, rill network
and distribution, as well as the magnitude of water flow associated with
transport and development of rill erosion.
To date, rill erosion studies have focus on rill formation over slopes of
constant gradient (Armstrong et al., 2011; Favis-Mortlock, 1998; Gessesse et al.,
2010; Shen et al., 2016). However, rill development and morphology in regions
where the gradient changes along the slope, have yet to be assessed. This works
intends to address this research gap by evaluating the influence of gradient
change and runoff volumes on rill erosion process, using a soil flume (3·9 × 1·4
m). In addition, morphological rill parameters will be estimated to allow a better
understanding of the rill erosion behaviour under different treatments.
2.2 Materials and Methods
2.2.1 Experimental design
All experiments were conducted in the soil erosion laboratory of the
Lancaster Environment Centre, Lancaster University, UK. A soil flume (3·9 m ×
34
1·4 m) was used to evaluate the relationship between slope and water flow in rill
formation. The flume was divided into two regions of different slope in order to
assess sedimentation and rill formation under varying gradients and surface
runoff. In the region A (see Figure 1), located 0 m to 1·5 m from the top of the
slope, the slope was either 6% or 9%, while in region B, the slope had a constant
value of 2% for both slopes (Figure 1). Whilst forming the surface, the slope
gradient was measured using a handheld clinometer (Suunto PM-5, accuracy ±
0·44%).
Fig. 1. Representative soil profile of the experimental plot determined from the
DEM (see section 2.2.4) showing the changing slope gradient and the position of
theta probe sensors along the ramp length for the 9% slope treatment. Area A:
variable slope gradient (6% or 9%); Area B: constant slope gradient.
Water was supplied to the top of the flume using a weir maintained at a
constant head. Two water flow rates were used (either 7 L min-1 or 12 L min-1) for
30 minutes. This gave four slope-discharge treatments (Table 1) which were
replicated in triplicate. To avoid edge effects, data from within 20 cm of the flume
edges were ignored.
35
Table 1. Combinations between water flow rates and slope gradient used on
experimental design.
Water flow rate
(L min-1)
Variable
slope
(%)
7 6
9
12 6
9
2.2.2 Soil flume set-up
The plot was filled with a 40 cm layer of sand underneath 30 cm of soil
(Fig. 2). The soil had 2.9% organic matter, a sandy loam texture with 5.1% clay,
43.6% silt and 51.3% sand. Soil aggregate size was standardized by sieving to 10
mm. The soil was added to the flume incrementally and compacted to a soil
density of 1·3 g cm-3. To maintain the same conditions for every experiment, the
soil surface was moistened until it reached field capacity. To measure soil
moisture, theta probe sensors were installed at depths of 0.07 m and 0.12 m, and
at distances of 0·5 m, 1·0 m, 2·0 m and 2·5 m from the top of the plot (Figure 1).
After each repetition, 5 cm of soil was removed from the surface layer and
replaced with new soil compacted to the same initial density. To ensure the same
subsurface conditions between slope angles, after each experimental run the
subsoil was turned over, raked and replaced.
36
Fig. 2. Soil flume filled with layers of sand (a) followed by soil (b).
2.2.3 Design of image acquisition system
Time-lapse photography was used to study the development of rill
erosion during runoff. Photographs were obtained before the water flow, during
flow at time intervals of one minute between each image, and 30 minutes after
the end of the experiment, which was the time required for drainage of excess
water at the soil surface. The images were captured using six synchronised
Canon digital SLR cameras (five EOS 450D and one Canon EOS 600D), with 28
mm fixed-focal lenses, located approximately 2·5 m above the soil surface. The
EOS 450D features a 12 mega-pixel CMOS sensor of 22·2 × 14·8 mm (a pixel size
of 5·2 μm) and delivered images of 4272 × 2848 pixels with a nominal ground
sampling distance of 0.5 mm. The EOS 600D features an 18 mega-pixel CMOS
sensor of 22·3 × 14·9 mm (a pixel size of 4·3 μm) and produced images of 5184 ×
3456 pixels with a nominal ground sampling distance of 0.4 mm. Lens apertures
were set to f/5·6 and, to maintain a fixed geometry during the experiment after
focussing, focus control was set to manual and adhesive tape was used to lock
37
the mechanisms. The cameras were arranged to produce three overlapping
convergent stereo-pairs over the plot (Fig. 3) with synchronous image acquisition
controlled by an intervalometer.
Fig. 3. Convergent stereo-pair cameras directed to the central portion of the plot
for time-lapse monitoring of rill erosion during runoff.
2.2.3.1 Flow velocity
To calculate the flow velocity, after each experimental run, red dye was
injected into the flow and photographs taken using an extra camera (Canon 500D,
15 mega-pixel, 22·3 × 14·9 mm CMOS sensor with 4·7 μm of pixel size, 4752 × 3168
pixels, and a zoom lens set to 18 mm, with aperture of f/5·6) at one image per
second for 10 seconds: the time required for the runoff to reach the end of the
plot (Figure 4).
38
Fig. 4. Dye-pigmented water flow used as tracer for flow velocity calculations. At
the sides of the experimental slope, photogrammetric control points for
georeferencing the cloud of points can be seen.
2.2.4 Photogrammetric workflow
The structure-from-motion (SfM) photogrammetry technique was used to
generate the three-dimensional (3D) point clouds. The images were processed
39
using Agisoft Photoscan v1·4 software, as already used in many studies
(Nouwakpo et al., 2014; Piermattei et al., 2016; Di Stefano et al., 2017; Prosdocimi
et al., 2017).
To generate the dense point cloud and then the digital elevation models,
the first step was image alignment. In this step, all images were processed in
order to detect the 2D location of matching tie point features in the images. For
this process, Photoscan uses custom algorithms that are similar to the Lowe’s
(2004) Scale Invariant Feature Transform (SIFT). The next step calculates camera
position and 3D location (X, Y and Z) of tie points by means of a bundle-
adjustment algorithm. As a result of these first two steps, a sparse 3D point cloud
was generated and then manually cleaned through removal of outliers to reduce
reconstruction errors.
For georeferencing, 14 ground control points (GCP) were located around
the plot (Fig. 4), with eight points used for control, and six as check points to
estimate precision and accuracy of the 3D models through the computation of
root-mean-square-error (RMSE). Also, the control points were used in the bundle
adjustment, ‘optimization’ in Photoscan. This process removes non-linear
distortions and minimises the total residual error on image observations by
simultaneously adjusting camera parameters and orientations, and the 3D point
positions (Granshaw, 1980). The point coordinates were established by total
station (Trimble C3, accuracy 2 mm), within an arbitrary local coordinate system.
The third step uses the camera locations estimated previously, to produce
a dense point cloud using multi-view reconstruction. The dense point clouds
40
were exported into Surfer software, converted to raster DEMs of 1-mm grid size
using the Kriging interpolation method, and cropped to remove the flume edges.
The photogrammetric parameters applied on Photoscan in the above-mentioned
steps are listed in Table 2.
Table 2. Photoscan parameters settings used during the point cloud generation.
Point cloud: alignment parameters Setting Accuracy Highest
Generic preselection No Reference preselection No Key point limit 0 Tie point limit 0 Filter point by mask No
Dense point cloud: reconstruction parameters
Quality Ultra-high
Depth filtering Mild
2.2.5 DEM of difference
DEMs of difference (DoD) were calculated to detect changes in the soil
surface topography over time and to spatially quantify the volumes of sediment
that were eroded and deposited. Georeferenced DEMs from different time
periods were subtracted from each other to produce a raster of morphological
change:
DoD = DEMt2 − DEMt1 (1)
where t1 is the initial time and t2 is the consecutive time of DEM acquisition.
Positive and negative values in the DoDs show deposition and erosion
respectively.
41
2.2.6 DEM uncertainty
DEM uncertainty was assessed through the generation of precision
estimates based on a Monte Carlo approach (James et al., 2017) with post-
processing tools in sfm_georef software (James and Robson, 2012). This method
consists of repeated bundle adjustments in Photoscan, in which different pseudo-
random offsets are applied to the image observations and the control
measurements to simulate observation measurement precision. Precision
estimates for each optimised model parameter were then derived by
characterising the variance for each particular parameter in the outputs from the
large number of adjustments. In this study, 4,000 bundle adjustments were
carried out, as used by James et al. (2017).
Precision maps were generated through the interpolation (1-mm grid size)
of the vertical standard deviation (σZ) derived by the precision estimates, to
enable precision estimates for both DEMs to be propagated into the DoD as
vertical uncertainties (Taylor, 1997; Brasington et al., 2003; Lane et al., 2003;
Wheaton et al., 2010). A spatially varying ‘level of detection’ (LoD) of significant
elevation change was calculated for each DoD cell, according to the equation:
LoD = t(σZ12 + σZ2
2)1
2⁄ (2)
where σZ1 and σZ2 are the vertical precision estimates for each cell in the two
DEMs and t is the t-distribution value defined by a specific confidence level (this
42
study 95%, giving t = 1·96). Thus, changes smaller than the LoD can be
disregarded, and Surfer was used to generate the LoD-thresholded DoD maps.
2.2.7 Rill development morphological indicators
The rills were classified manually in the DEMs. The morphological
indicators chosen were total rill length (RL, m), mean rill width (�̅�, cm), mean
rill depth (�̅�, cm), rill width-depth ratio (WD) and rill density (RD, m m-2). The
indicators were calculated according to the equations:
RL = ∑ RLini=1 (3)
where RLi is the i rill length (m), and n is the number of rills,
�̅� = ∑ 𝑤𝑖
𝑛𝑖=1
𝑛 (4)
where wi is the i rill width (cm),
�̅� = ∑ 𝐷𝑖
𝑛𝑖=1
𝑛 (5)
where Di is the i rill depth (cm), and
𝑊𝐷 =�̅�
�̅� (6).
43
Rill density is a measure of the rill erosion coverage in the area, being directly
proportional to soil erosion and bifurcation ratio, and also reflect the degree of
rill fragmentation (Shen et al., 2015).
𝑅𝐷 =∑ 𝑅𝐿𝑛
𝑖=1
𝐴 (7)
where A is the study area (m2).
The erosion measurements were performed using the Simpson's rule
method (see Easa, 1988), which assumes nonlinearity in the profile between grid
points. This technique shows greater precision in the determination of volume
compared to linear methods, such as the trapezoidal rule (Fawzy, 2015). The soil
volume was converted to mass (kg) through soil bulk density.
Principal component analysis (PCA) was used to assess the relationship
between morphological and quantitative indicators of rill erosion. Each variable
was standardized to have mean zero and unity variance, to avoid the effects of
differences in scales or magnitudes of the variables. The first two principal
components were visualized in a PCA biplot to represent how the variables relate
to one another and how the observations differ regarding to those variables.
2.3 Results
2.3.1 Precision results
44
The photogrammetric errors (RMSE) calculated by the Photoscan on x,y
and z-axes for the control, check and tie points of each SfM point cloud are listed
in Table 3. The point clouds show average errors of order ~1 mm on xyz on control
and check points, whereas the tie points image residual RMS was ~ 0·3 pix.
Table 3. Root mean square error (RMSE) of check points, control points and tie
points image residuals.
Water flow
(L min-1)
Slope
(%)
RMS tie points
image
residuals (pix)
RMSE of control points
(mm)
RMSE of check points
(mm)
X Y Z X Y Z
7
6
0·35 0·872 0·468 1·323 0·775 1·742 0·592
0·32 0·619 0·505 0·918 0·536 1·721 1·372
0·28 0·651 0·832 0·876 0·291 0·301 0·419
Average 0·32 0·714 0·602 1·039 0·534 1·255 0·794
9 0·31 0·49 0·719 0·722 0·789 0·022 0·825
0·27 0·668 0·624 0·572 0·378 0·084 0·424
Average 0·29 0·579 0·672 0·647 0·584 0·053 0·625
12
6
0·32 0·604 0·981 0·616 0·408 0·841 0·752
0·28 0·33 0·565 0·681 0·517 0·386 0·389
0·31 1 0·887 0·77 1·052 0·533 0·89
Average 0·30 0·645 0·811 0·689 0·659 0·587 0·677
9
0·47 0·777 0·674 1·456 0·597 0·241 0·237
0·36 0·804 0·728 0·673 0·679 0·092 0·682
0·35 0·641 0·485 0·742 0·052 1·197 0·578
Average 0·39 0·741 0·629 0·957 0·443 0·510 0·499
The precision maps show the spatial variation of precision along the flume
(Fig. 5), with LoD values ranging from 1 mm to 2·8 mm. The smaller values were
concentrated in the area around the centroid of control.
45
Fig. 5. Digital elevation models, precision maps and LoD showing the spatial
distribution of the error along the flume. Changes with magnitudes smaller than
the LoD can be disregarded.
2.3.2 DEM of Difference (DoD)
The DoD maps obtained for the flume (Figs. 6 and 7) show the different
spatial behaviour of rill erosion due to changes in the slope and water flow rate.
Both erosion (red), and deposition (blue) were detected. The treatments with 7 L
min-1 of flow rate (Fig. 6) showed short and shallow rills, whereas the rills from
12 L min-1 of water flow rate were longer and deeper (Fig. 7).
The sediment depositional area corresponded with the gradient change
for 7 L min-1 treatments, with rills on the steeper section of the flume and the
depositional areas on the shallower section (Fig. 6). However, at 12 L min-1, the
46
water flow was able to transport the soil further past the gradient change, causing
deposition at the bottom of the flume and higher rates of rill erosion.
Fig. 6. DEM of difference (DoD) maps, overlain over hillshaded topography,
showing rill erosion over runoff at 7 L min-1 of flow rate in two slopes, 6 % (A)
and 9 % (B). Colour scale ranges from red (erosion) to blue (deposition).
47
Transparent regions mean no significant changes (DoD is less than the level of
detection).
Fig. 7. DEM of difference (DoD) maps, overlain over hillshaded topography,
showing rill erosion over runoff at 12 L min-1 of flow rate in two slopes, 6% (A)
48
and 9% (B). Colour scale ranges from red (erosion) to blue (deposition).
Transparent regions mean no significant changes (DoD is less than the level of
detection).
2.3.3 Rill erosion and soil loss
The rates of soil loss and rill erosion were greater at 12 L min-1 than at 7 L
min-1 of water flow, regardless the slopes (Table 4). Increasing the water flow rate
from 7 L min-1 to 12 L m-1 increased the amount of soil removed by rill erosion by
about three times, on both slopes assessed. For the water flow rate of 7 L min-1,
the steeper-slope experiments showed greater values of soil erosion than the
shallower-slope experiments, whereas for 12 L min-1 the amount of soil loss was
about 1 kg m-2 greater at 6% than at 9% of slope. The proportion of rill erosion as
a fraction of the total soil loss increased with the water flow and slope. However,
the water flow rate contributed the most to the formation of rill erosion. In the
treatments with 12 L min-1, the rill erosion accounted for more than a half of the
soil loss in the study area.
Table 4. Rill erosion and soil loss in different water flows and slopes.
Water flow
(L min-1)
Slope
(%)
Rill erosion
(kg m-2)
Soil loss
(kg m-2)
Proportion of soil loss due
to rill erosion
(%)
7 6 0·58 d 2·32 d 25·1
9 0·94 c 2·77 c 33·9
12 6 2·42 b 4·12 a 58·8
9 3·05 a 3·72 b 82·0
49
Means in a column followed by the same letter are not significantly different from
one another (α = 0·05) according to the Tukey test.
2.3.4 Rill morphology
Total rill length increased with an increase slope at 7 L min-1, but at 12 L
min-1, the slope had an opposite effect on rill length (Fig. 8). The total rill length
on 12 L min-1 at 6% of slope was 2.1 times longer than the 9% slope. Otherwise,
the main rill length was longer on low slopes for both water flow rates (Fig. 8).
50
Fig. 8. Rill erosion characteristics and morphological parameters under different
water flow and slope gradients. Error bars represent the standard error of the
mean (n = 3).
The mean rill depth showed a positive relationship with the water flow
rate and slope. On the other hand, the rills were wider on low flow rates and
51
slopes, with the 7 L min-1 at 6% slope showing the widest rill among the
treatments studied (Fig. 9). This is reflected in the behaviour of the rill width-
depth ratio (WD). Taking the WD ratio value for 12 L min-1 flow rate as the
example, at 9% slope the WD ratio was approximately four times lower than the
7 L min-1 flow rate at 6% slope.
Fig. 9. Principal component analysis (PCA) for the morphological and
quantitative indicators of rill erosion in flows of 7 and 12 L m-1 and slopes of 6%
and 9%. FV: flow velocity; RD: rill density; MD: mean rill depth; MRL: main rill
length; MW: mean rill width; WD: width-depth ratio; RE: rill erosion; SL: soil
loss.
52
The flow velocity (FV) for 7 L min-1 flow rate did not show a significant
difference between slopes of 6% and 9%. On the other hand, the FV increased by
1·28 times with the increase of slope for 12 L min-1 flow rate. At 6% the change in
water flow rate did not affect the FV, whereas, when considering 9 % slope, the
increase from 7 L min-1 to 12 L min-1 increased the FV by 1·32 times.
The rill density showed a different behaviour according to the rate of the
water flow. For 7 L min-1, the rill density increased with the increase of slope.
However, for 12 L min-1, the density of rills in the area decreased when the slope
increased, being 2·08 times higher in 12 L min-1 at 6% than at 9% slope.
2.4 Discussion
2.4.1 Rill network development
In this study we introduce the use of close-range photogrammetry to
study the rill network development with millimetric accuracy, allowing the
extraction of rills and parameter calculations (i.e. rill width, depth, length) by
applying image processing methods. The spatial distribution of the rills (Figs. 6
and 7) shows that the highest rates of soil erosion were at the top of the plot, with
erosion decreasing down-slope. This occurs mainly because of the gradient
change, with slope decreasing and becoming constant at the bottom of the soil
flume. In addition, the clean water at the top of the rills has the maximum soil
detachment potential (Nearing et al., 1989; Chen et al., 2014). Also, as the
53
concentration of the sediments on the water flow increases along the rills, the soil
detachment capability continues to decrease, due to the energy being used by the
transportation of soil particles (Chen et al., 2016).
Table 4 shows that the amount of soil eroded by rill erosion increased with
flow rates and slope gradients. The increase in the slope gradient also promotes
an increase of the velocity and energy of concentrated flow (Li et al., 2005).
Similar results were founded by Chen et al. (2016) and Kinnell (2000).
Furthermore, another parameter that influenced rill erosion was FV,
which has an important role in rill development and soil erosion. When the flow
rate was at its highest level (12 L min-1) the increase in slope gradient increased
the FV, leading to a greater eroding force. Thus, higher FV values imply the
generation of deeper rills, because it is easier for the water to erode an existing
rill than create another one (Mancilla et al., 2005). Then, the rill erosion intensity
was closely associated with the FV in the area.
The slope gradient change along the plot showed clearly that the sediment
transport capacity was directly related to changes in flow velocity (Lei et al.,
1998). For the two different discharges, sediment deposition was different in the
2% slope area. For the 7 L min-1, the transported sediments were deposited in thin
layers at the boundary between the area of the higher and shallower slope.
However, for the 12 L min-1 flow rate, deposition was concentrated towards the
bottom of the rills, further into the lower slope area. This reflects the greater
distance required for the flow energy to fall to the point at which deposition
occurs.
54
2.4.2 Soil loss due to rill erosion
The proportion of soil loss due to rill erosion was directly affected by the
increase of slope and water flow rate. Studies show that the increase of slope
gradient also increase the collapse of rill heads and sidewalls, accounting for > 90
% of rill erosion (Xiao et al., 2016), which can lead to a greater contribution to
total soil loss (Jiang et al., 2018; Wang and Shi, 2015). Zheng et al. (1987) and Shen
et al. (2015), also observed that rill erosion accounted for up to 74·2 % and 86·7 %
of soil losses, respectively.
Contrary to expectations, for the 12 L min-1 flow rate experiments, the
increase of slope gradient did not coincide with an increase in soil loss. This can
be explained by the higher bifurcation showed for 6% slopes in contrast with for
9% slopes (Fig. 7). On steeper slopes, the rill network pattern comprises deep rills
with fewer secondary channels, whereas at 6% slope, the rills were shallower and
with high values of rill density, producing greater total soil losses but smaller rill
erosion rates (Fig. 9).
When comparing different types of soil erosion, rill erosion generates
greater soil losses than other types of soil erosion (sheet and splash erosion) (Shen
et al., 2015). Thus, once rill erosion is initiated and became the main erosion
pattern, the rate of soil loss increases very rapidly (Kimaro et al., 2008; He et al.,
2016).
2.4.3 Rill network morphology
55
As the flow velocity increase, the rills became deeper and narrow (Fig. 9).
Mean rill width decreased when the slope gradient increased, regardless of the
water flow rate, while the mean rill depth increased with increasing slope.
Similar results were founded by Lei and Nearing (2000) and He et al. (2016). This
behaviour can be explained by the increase in eroding force and erosion intensity
promoted by the high stream power on high slopes (Chen et al., 2016).
Rill density values reflected the rill fragmentation (Figs. 6 and 7) and were
related to the main rill length (Fig. 9). The highest values of rill density were for
12 L min-1 at 6 % of slope. On the other hand, the rill density was statistically the
same for 7 L min-1 at 6% and 12 L min-1 at 9% of slope. This means that high water
flow, combined with high slope gradient can generate substantial amounts of soil
loss in non-complex rill network. Thus, the amount of rill erosion is primarily
related to the depth and flow velocity than the rill network density.
2.5 Conclusions
In this study, a runoff simulator was used to investigate the influence of
different slopes and water flow on the behaviour of soil erosion and sediment
transport. The gradient change had a great influence on the rill erosion behaviour
along the slope, with the highest soil erosion rates being concentrated in the
upper part of the plot. The gradient change also resulted in different patterns of
rill network, reducing depth and promoting soil sedimentation. The flow velocity
in rills increased with water flow and slope, showing a strong correlation with
56
the amount of rill erosion. The rills produced at a lower water flow rate (7 L min-
1) were wider, with width-depth ratio almost 5 times greater at 6% slope than at
12 L min-1 at 9% of slope.
The greatest amount of soil loss and rill density was found for 12 L min-1
on a 6% slope, whereas the greatest rill erosion was on 12 L min-1 at 9% of slope.
These results demonstrate that on steep slopes the soil erosion is dominated by
the rill erosion with less rill network density while, on low slopes, there are other
types of soil erosion occurring together with rill erosion, causing the reduction of
soil loss due to rill erosion.
Despite the known effects of vegetation cover on erosion reduction and
sediment production, this work did not evaluate treatments with vegetation on
the surface. Obtaining high-resolution DEMs through SfM in vegetated areas
requires further processing, using mathematical algorithms to classify and filter
the point cloud in vegetated and non-vegetated areas. Thus, the accuracy of
DEMs obtained by SfM in areas with vegetation cover is compromised, as well
as volumetric measurements in these regions. The use of SfM to study the
morphology and development of the furrow network in vegetated areas is still a
challenge.
2.6 References
Abd Elbasit, M.A.M., Anyoji, H., Yasuda, H., Yanamoto, S., 2009. Potential of low cost
close-range photogrammetry system in soil microtopography quantification. Hydrol.
Process. 23, 1408–1417.
57
Aguilar, M.A., Aguilar, F.J., Negreiros, J., 2009. Off-the-shelf laser scanning and close-
range digital photogrammetry for measuring agricultural soils microrelief. Biosyst. Eng.
103, 504–517.
Armstrong, A., Quinton, J.N., Heng, B.C.P., Chadler, J.H., 2011. Variability of interrill
erosion at low slopes. Earth Surf. Process. Landforms 36, 97-106.
Bewket, W., Sterk, G., 2003. Assessment of soil erosion in cultivated fields using a survey
methodology for rills in the Chemoga watershed, Ethiopia. Agric. Ecosyst. Environ. 97,
81–93.
Brasington, J., Langham, J., Rumsby, B., 2003. Methodological sensitivity of
morphometric estimates of coarse fluvial sediment transport. Geomorphology 53, 299–
316.
Brodu, N., Lague, D., 2012. 3D terrestrial Lidar data classification of complex natural
scenes using a multi-scale dimensionality criterion: applications in geomorphology.
ISPRS J. Photogramm. Remote Sens. 68, 121–134
Chen, X.Y., Zhao, Y., Mo, B., Mi, H.X., 2014. An improved experimental method for
simulating erosion processes by concentrated channel flow. Plos One 9 (6), e99660.
Chen, X.Y., Zhao, Y., Mi, H., Mo, B., 2016. Estimating rill erosion process from eroded
morphology in flume experiments by volume replacement method. Catena 136, 135-140.
Di Stefano, C., Ferro, V., Palmeri, V., Pampalone, V., 2017. Measuring rill erosion using
structure from motion: A plot experiment. Catena 156, 383-392.
Easa, S.M., 1988. Estimating pit excavation volume using nonlinear ground profile. J.
Surv. Eng. 114, 71–83.
Eltner, A., Maas, H-G., Faust, D., 2018. Soil micro-topography change detection at
hillslopes in fragile Mediterranean landscapes. Geoderma 313, 217-232.
Favis-Mortlock, D.T., Boardman, J., Parsons, A.J., Lascelles, B., 2000. Emergence and
erosion: a model for rill initiation and development. Hydrol. Process. 14, 2173–2205.
Fawzy, H.E-D., 2015. The accuracy of determining the volumes using close range
photogrammetry. J. Mech. Civ. Eng. 12, 10–15.
Gatto, L.W., 2000. Soil freeze–thaw-induced changes to a simulated rill: potential
impacts on soil erosion. Geomorphology 32, 147–160.
Gessesse, G.D., Fuchs, H., Mansberger, R., Klik, A., Rieke-Zapp, D.H., 2010. Assessment
of erosion, deposition and rill development on irregular soil surfaces using close range
digital photogrammetry. Photogramm. Rec. 25, 299–318.
Govers, G., Poesen, J., 1988. Assessment of the interrill and rill contributions to total soil
loss form an upland field plot. Geomorphology 1, 343–354.
58
Guo, M., Shi, H., Zhao, J., Liu, P., Welbourne, D., Lin, Q., 2016. Digital close range
photogrammetry for the study of rill development at flume scale. Catena 143, 265–274.
Granshaw, S.I., 1980. Bundle adjustment methods in engineering photogrammetry.
Photogramm. Rec. 10 (56), 181–207.
He, J.J., Sun, L.Y., Gong, H.L., Cai, Q.G., Jia, L.J., 2016. The characteristics of rill
development and their effects on runoff and sediment yield under different slope
gradients. J. Mt. Sci. 13, 397–404.
Heng, B.C.P., Chandler, J.H., Armstrong, A., 2010. Applying close-range digital
photogrammetry and soil erosion studies. The Photogrammetric Record 25, 240–265.
Jackson, J.A., 1997. Glossary of Geology. American Geological Institute, Alexandria.
James, M.R., Robson, S., 2012. Straightforward reconstruction of 3D surfaces and
topography with a camera: Accuracy and geoscience application. J. Geophys. Res. 117,
F03017.
James, M.R., Robson, S., Smith, M.W., 2017. 3-D uncertainty-based topographic change
detection with structure-from-motion photogrammetry: precision maps for ground
control and directly georeferenced surveys. Earth Surf. Process. Landforms 42, 1769-
1788.
Jiang, F., Zhan, Z., Chen, J., Lin, J., Wang, M.K., Ge, H., Huang, Y., 2018. Rill erosion
processes on a steep colluvial deposit slope under heavy rainfall in flume experiments
with artificial rain. Catena 169, 46-58.
Kimaro, D.N., Poesen, J., Msanya, B.M., Deckers, J.A., 2008. Magnitude of soil erosion on
the northern slope of the Uluguru Mountains, Tanzania: interrill and rill erosion. Catena
75, 38–44.
Kinnell, P.I.A., 2000. The effect of slope length on sediment concentrations associated
with side-slope erosion. Soil Sci. Soc. Am. J. 64, 1004–1008.
Lane, S., Westaway, R., Murray Hicks, D., 2003. Estimation of erosion and deposition
volumes in a large, gravel-bed, Braided River using synoptic remote sensing. Earth Surf.
Process. Landf. 28, 249–271.
Lei, T.W., Nearing, M.A., 1998. Rill erosion and morphological evolution: a simulation
model. Water Resour. Res. 34, 3157–3168.
Lei T.W., Nearing, M.A., 2000. Flume experiments for determining rill hydraulic
characteristic erosion and rill patterns. J. Hydraul. Eng. 31, 49-55.
Lei, T.W., Nearing, M.A., Haghighi, K., 1998. Rill erosion and morphological evolution:
a simulation model. Water Resour. Res. 34, 3157–3168.
Li, P., Li, Z.B., Zheng, L.Y., Lu, K.X., 2005. Comparisons of dynamic mechanics of soil
erosion and sediment yield by runoff on loess slope. Soil Water Conserv. 19, 66–69.
59
Lisein, J., Linchant, J., Lejeune, P., Bouché, P., Vermeulen, C., 2013. Aerial surveys using
an unmanned aerial system (UAS): comparison of different methods for estimating the
surface area of sampling strips. Trop. Conserv. Sci. 6, 506–520.
Lowe D.G., 2004. Distinctive image features from scale-invariant keypoints. Int. J.
Comput. Vis. 60, 91–110.
Mancilla, G.A., Chen, S., McCool, D.K., 2005. Rill density prediction and flow velocity
distributions on agricultural areas in the Pacific Northwest. Soil Tillage Res. 84, 54–66.
distributions on agricultural areas in the Pacific Northwest. Soil Tillage Res. 84, 54–66
Morgan, R.P.C., 1977. Soil Erosion in the United Kingdom: Field studies in the Silsoe
area, 1973–1975. Occasional Paper. 4 Nat. Coll. Agr. Eng. Silsoe, UK.
Morgan, R.P.C., 2005. Soil Erosion and Conservation. third ed. Blackwell Publishing
Ltd., Cornwall.
Nearing, M.A., Foster, G.R., Lane, L.J., 1989. A process based soil erosion model for
USDA Water Erosion Prediction Project Technology. Trans. ASAE 32, 1587–1593.
Nearing, M.A., Norton, L.D., Bulgakov, D.A., Larionov, G.A., West, L.T., Dontsova,
K.M., 1997. Hydraulics and erosion in eroding rill. Water Resour. Res. 33, 865–876.
Nouwakpo, S.K., Huang, C.-H., 2012. A simplified close-range photogrammetric
technique for soil erosion assessment. Soil Sci. Soc. Am. J. 76, 70–84.
Nouwakpo, S.K., James, M.R., Weltz, M.A., Huang, C.-H., Chagas, I., Lima, L., 2014.
Evaluation of structure from motion for soil microtopography measurement.
Photogramm. Rec. 29, 297–316.
Piermattei, L., Carturan, L., de Blasi, F., Tarolli, P., Dalla Fontana, G., Vettore, A., Pfeifer,
N., 2016. Suitability of ground-based SfM–MVS for monitoring glacial and periglacial
processes. Earth Surf. Dyn. 4, 425–443.
Poesen, J., Nachtergaele, J., Verstraeten, G., Valentin, C., 2003. Gully erosion and
environmental change: importance and research needs. Catena 50, 91–133.
Prosdocimi, M., Burguet, M., Di Prima, S., Sofia, G., Terol, E., Rodrigo Comino, J., Cerdá,
A., Tarolli, P., 2017. Rainfall simulation and structure-from-motion photogrammetry for
the analysis of soil water erosion in Mediterranean vineyards. Sci. Total Environ. 574,
204–215.
Rieke-Zapp, D.H., Nearing, M.A., 2005. Digital close range photogrammetry for
measurement of soil erosion. Photogramm. Rec. 20, 69–87.
Robertson, D.P., Cipolla, R., 2009. Structure from motion. In: Varga, M. (Ed.), Practical
Image Processing and Computer Vision. John Wiley & Sons, Chichester.
Shen, H., Zheng, F., Wen, L., Han, Y., Hu, W., 2016. Impacts of rainfall intensity and
slope gradient on rill erosion processes at loessial hillslope. Soil Tillage Res. 155, 429-436.
60
Shen, H.O., Zheng, F.L., Wen, L.L., Lu, J., Jiang, Y.L., 2015. An experimental study of rill
erosion and morphology. Geomorphology 231, 193–201.
Taylor, J.R., 1997. An Introduction to Error Analysis: the Study of Uncertainties in
Physical Measurements, second edition. University Science Books: Sausalito, California.
Wang, L., Shi, Z.H., 2015. Size selectivity of eroded sediment associated with soil texture
on steep slopes. Soil Sci. Soc. Am. J. 79, 917–929.
Wheaton, J., Brasington, J., Darby, S., Sear, D., 2010. Accounting for uncertainty in DEMs
from repeat topographic surveys: improved sediment budgets. Earth Surf. Process.
Landf. 35, 136–156.
Wirtz, S., Seeger, M., Ries, J.B., 2012. Field experiments for understanding and
quantification of rill erosion processes. Catena 91, 21–34.
Xiao, H., Liu, G., Liu, P.L., 2016. Response of detachment rate of loess slope to
hydrodynamic characteristics under concentrate flow condition. Trans. CSAE 32, 106–
111 (in Chinese with English abstract).
Zhang, W., Qi, J., Wan, P., Wang, H., Xie, D., Wang, X., Yan, G., 2016. An Easy-to-Use
Airborne LiDAR Data Filtering Method Based on Cloth Simulation. Remote Sens. 8, 501.
Zheng, F.L., Tang, K.L., Zhou, P.H., 1987. Approach to the genesis and development of
rill erosion on slope land and the way to control. J. Soil Water Conserv. 1, 36–48 (in
Chinese, with English Abstract).
61
3. High resolution monitoring of sheet (interrill) erosion using structure-
from-motion
3.1 Introduction
Soil erosion is one of the main factors that lead to the degradation of
agricultural land worldwide (Boardman et al., 2003; Bakker et al., 2004; Zhao et
al., 2019). It threats agricultural sustainability by reducing the water retention
capacity, the nutrient content, and total organic carbon of the soil (Quinton et al.,
2010; Zhao et al., 2016), causing pollution of water bodies (Lal, 1998). Thus, the
accurate measurement of erosion rates becomes a key factor for better
understanding the erosive process in different scenarios and to promote efficient
recovery strategies aiming to reduce soil loss in sloping areas (Cerdan et al., 2010;
Di Stefano and Ferro, 2017).
In the last decades, soil erosion has been studied primarily from plots of
soil loss under natural and artificial rainfall conditions (Araya et al., 2011; García-
Ruiz et al., 2015; Guo et al., 2015; Fang et al., 2017; Zhao et al., 2019). However,
soil erosion data acquired from traditional methods is considered a time-
consuming and costly process (Cerdan et al., 2010). Recently, with the
technological advances, digital elevation models (DEM) produced from high-
resolution surveying techniques have played an important role in the
understanding of geomorphological processes. These advances have been
facilitated by the development of Structure-from-Motion (SfM; Ullman, 1979), a
62
technique that combines well-established photogrammetric principles with
modern computational methods (Westoby et al., 2012).
SfM photogrammetry, using images acquired from unmanned aerial
vehicle (UAV), is being adopted for the generation of high-resolution DEMs in
studies of surface processes (Colomina and Molina, 2014). The use of UAVs have
made the acquisition of aerial photographs cheap and easy, allowing surveys at
high temporal and spatial resolution. This makes it possible to monitor and
quantify rapidly changing landscapes (Cook, 2017).
In geosciences, the application of photogrammetry using SfM is now
considered an established method to describe high-resolution topography (Cook,
2017; Eltner et al., 2018). This technique has been used in several Earth surface
surveys, in studies of fluvial, glacial, and coastal geomorphological processes
(Dietrich, 2016; Westoby et al., 2016; Warrick et al., 2017), as well as in the
monitoring and quantification of soil erosion volumes in gullies (Castillo et al.,
2012; Gómez- Gutiérrez et al., 2014; Stöcker et al., 2015, Glendell et al., 2017).
However, photogrammetry applications from SfM with the use of UAVs
in studies of soil erosion where there are no large mass movements and gullies
are still scarce. This is due to difficulties in defining a stable coordinate reference
system, which is important for quantifying changes of small magnitudes that are
typical of laminar erosion processes. There are studies involving SfM and UAVs
that quantify erosion at large magnitudes (Bazzoffi, 2015, Neugirg et al., 2016);
however, the evaluation of sheet erosion under natural rainfall is still limited
63
(Eltner et al., 2015; Nouwakpo et al., 2015, Hänsel et al., 2016, Morgan et al., 2017;
Prosdocimi et al., 2017).
The assessment of the accuracy of data derived from SfM has been carried
out by several studies (James and Robson, 2012; Westoby et al., 2012; Gómez-
Gutiérrez et al., 2014; Eltner et al., 2015; Cook et al., 2017; James et al., 2017a;
Morgan et al., 2017) using aerial and terrestrial laser scanning or control points
with high precision as a reference. The reported accuracies vary widely from sub-
decimetre to more than 1 m, reflecting the dependence of SfM accuracy on the
image quality, distortion and orientation, vegetation presence, soil surface
characteristics, number and precision of the ground control points and scale.
The SfM relative precision ratio (measurement precision: observation
distance) is limited by the model used for camera calibration in software, but
generally exceed 1:1000, which implies the accuracy of centimetres over distances
of 10s of metres. (James and Robson, 2012). Comparative studies have shown that
SfM generates topographic data with quality, resolution and accuracy similar to
those obtained by laser scanning and classical photogrammetry surveys
(Westoby et al., 2012; Fonstad et al., 2013; Mancini et al., 2013; Stumpf et al., 2013;
White et al., 2013; James and Quinton, 2014; Ouédraogo et al., 2014; Cook, 2017).
Repeated topographic surveys of the same area are often carried out in
order to establish spatial patterns of erosion, deposition, and changes in volume.
Therefore, when successive DEMs are subtracted from each other, the DEM of
difference (DoD) can be generated, highlighting areas of erosion and deposition
(Lane et al., 2003). However, there are no studies that validate such volume
64
measurements conducted by UAVs through SfM and DoD through comparison
with measurements of sediment collected in standard erosion plots.
The speed of data acquisition from UAVs, coupled with the high precision
of the DEMs generated by the SfM can be important tools in obtaining soil loss
values that are used in modelling water erosion. Thus, this study aims to evaluate
the efficacy of SfM in obtaining sub-centimetre level precision measurements of
soil loss in areas of sheet erosion under natural rainfall in standard plots of soil
loss where there are no large furrows or gullies.
3.2 Materials and Methods
3.2.1 Experimental area
All the experiments were conducted on the campus of the Federal
University of Lavras, Lavras, Brazil (21º13'20'' S and 44º58'17'' W), during two
hydrological years. The area presents a typical humid subtropical climate, with
an annual average rainfall of 1,530 mm. The soil is classified as an Inceptisol,
according to the US Soil Taxonomy, with 47·8% sand, 15·8% silt and 36·4% clay,
presenting a density of 1,400 kg m-3. Three plots (12 m × 4 m) were installed in
the area to monitor soil erosion on a 23% slope, under bare soil and natural
rainfall (Fig. 1). The longest dimension of the plot followed the direction of the
slope.
65
Fig. 1. Typical erosion plot showing dimensions and control point layout.
3.2.2 Sediments measurements on erosion plots
The collector system comprised two tanks installed in sequence, the first
with 500 L capacity and the second 250 L (Fig. 2). Among the sedimentation tanks
there was a Geib divisor system with 15 windows so that after filling the first
tank, only 1/15 of the runoff was conducted to the second tank.
66
Fig. 2. Runoff collection system used on soil loss plots. Inset shows the detail of
a ground control point.
To quantify soil losses, runoff samples and sediments were collected from
the collection tanks. After stirring, three aliquots of predetermined volume were
collected, transferred to the laboratory, the supernatant decanted and the
remaining sediment dried at 105°C before weighing.
3.2.3 Image acquisition
A UAV, DJI Phantom 3 Professional, was used for data acquisition. The
UAV features integrated a gimbal-stabilized FC300X camera with 12-megapixel
(4000 × 3000) Sony EXMOR 1/2·3 sensor, 94º field of view (FOV) and 20-mm focal
length. The lens aperture was set to f/2·8 and images acquired in RAW format.
67
Seven flights were performed on each erosion plot, from June 2016 to April
2018. The flights were conducted manually using a combination of orthogonal
and oblique photos to provide convergent image geometries between the lines
(James et al., 2014). In order to reduce the influence of direct sunlight at noon,
flights were conducted either in the morning or in the afternoon on cloudy days.
Flight heights were over 4 m with a nominal ground sampling distance of 1·5
mm. A total of 35 photos were taken in each survey, with 70% of forward and
side overlap.
For georeferencing, 14 ground control points (GCP) were installed around
the plots (Fig. 1), with ten points used for control and four as check points to
estimate the precision and the accuracy of the 3D models by calculating the root
mean square error (RMSE). The coordinates of the points were established by
total station (Geodetic GD2i, accuracy 2 mm), within an arbitrary local coordinate
system.
3.2.4 Structure from motion (SfM) point cloud generation
The generation of three-dimensional point clouds (3D) was performed
using the SfM photogrammetry technique, which allows the reconstruction of the
topography from randomly distributed and oriented images from uncalibrated
cameras (James and Robson, 2012; Fonstad et al., 2013; Agüera-Vega et al. 2018).
The images were processed using the commercially available SfM software
Agisoft Photoscan Professional® v1.4, which have been used for several studies
68
(Brunier et al., 2016; Di Stefano et al., 2017; Prosdocimi et al., 2017). All processing
was done through cloud computing using a virtual machine (24 Intel Xeon
Platinum 3.7 GHz CPUs, two NVIDIA Tesla K80 GPUs and 128 GB RAM).
Firstly, image alignment was done to generate the dense point cloud and
then the DEMs. In this step, all images were processed in order to detect the 2D
location of matching tie point features in the images. For this process, Photoscan
uses custom algorithms that are similar to the Lowe’s (2004) Scale Invariant
Feature Transform (SIFT). The next step calculates camera position and 3D
location (X, Y and Z) of tie points by means of a bundle-adjustment algorithm.
The control points were used in the bundle adjustment, ‘optimization’ in
Photoscan. This process reduces non-linear distortions and minimises the total
residual error on image observations by simultaneously adjusting camera
parameters and orientations, and the 3D point positions (Granshaw, 1980). As a
result of these first two steps, a sparse 3D point cloud was generated.
The third step uses the camera locations estimated previously, to produce
a dense point cloud using multi-view reconstruction. The dense point clouds
were exported into Surfer 16 software, converted to raster DEMs of 4-mm grid
size using the nearest neighbour interpolation method, and cropped to remove
the plot edges. The photogrammetric parameters applied on Photoscan in the
above-mentioned steps are listed in Table 1.
69
Table 1. Photoscan parameters settings used during the point cloud generation.
Point cloud: alignment parameters Setting Accuracy Highest
Generic preselection Yes Reference preselection Yes Key point limit 120,000 Tie point limit 0 Filter point by mask No
Dense point cloud: reconstruction parameters
Quality Medium
Depth filtering Mild
3.2.5 Erosion measurements using SfM
The erosion measurements in each plot were performed using the
Simpson's rule method (see Easa, 1988), which assumes nonlinearity in the
profile between grid points. This technique shows greater precision in the
determination of volume compared to linear methods, such as the trapezoidal
rule (Fawzy, 2015). The soil volume was converted to mass (kg) through soil bulk
density, to correlate with the sediment collected from each runoff tank in the
interval between the two drone flights.
DEMs of difference (DoD) were calculated to detect changes in the soil
surface topography over time and to spatially quantify the volumes of sediment
that were eroded and deposited. This technique consists of subtracting
georeferenced DEMs from different periods to generate a raster of morphological
change:
DoD = DEMt2 − DEMt1 (1)
70
where t1 is the initial time and t2 is the consecutive time of DEM
acquisition. Positive and negative values in the DoDs show deposition and
erosion respectively.
3.2.6 DEM uncertainty and Level of Detection (LoD)
DEM uncertainty was assessed through the generation of precision
estimates based on a Monte Carlo approach (James et al., 2017a) with post-
processing tools in sfm_georef software (James and Robson, 2012). This method
consists of repeated bundle adjustments in Photoscan, in which different pseudo-
random offsets are applied to the image observations and the control
measurements to simulate observation measurement precision. Precision
estimates for each optimised model parameter were then derived by
characterising the variance for each particular parameter in the outputs from the
large number of adjustments. In this study, 4,000 bundle adjustments were
carried out, as used by James et al. (2017a).
Precision maps were generated through the interpolation (4-mm grid size)
of the vertical standard deviation (σZ) derived by the precision estimates, to
enable precision estimates for both DEMs to be propagated into the DoD as
vertical uncertainties (Taylor, 1997; Wheaton et al., 2010). A spatially varying
‘level of detection’ (LoD) of significant elevation change was calculated for each
DoD cell, according to the equation:
LoD = t(σZ12 + σZ2
2)1
2⁄ (2)
71
where σZ1 and σZ2 are the vertical precision estimates for each cell in the
two DEMs and t is the t-distribution value defined by a specific confidence level
(this study 95%, giving t = 1·96). Thus, changes smaller than the LoD can be
disregarded, and Surfer was used to generate the LoD-thresholded DoD maps.
3.2.7 Statistical analysis
For assessing the agreement between measurements obtained by sediment
collection (response variable) and by SfM (explanatory variable) a linear
regression model was fit to the data. Because the same plots were repeatedly used
through time for data collection, we investigated whether measurements from
the same plot were statistically dependent by introducing a random intercept for
each plot in the linear regression model, following a mixed modelling approach
(Gelman and Hill, 2007; Zuur et al., 2009).
However, after fitting the model, we observed that the variance associated
with the random intercept was null, indicating no evidence of statistical
dependence caused by the plot effect. A drawback of that approach is the low
number (three) of groups available for estimating the variance associated with
the random effect of plots.
As an alternative approach to further investigate whether a statistical
dependence among observations could be attributed to a plot effect, an analysis
of covariance was performed, with both plot and SfM as explanatory variables,
and amount of collected sediments as response variable. In agreement with the
72
results from the previous approach, no significant effect of plots was observed
(F2,14 = 0.4, P = 0.68). For the above reasons, the final model was simplified by
omitting the plot effect and an ordinary linear regression approach was used,
assuming statistical independence of the model residuals.
3.3 Results
3.3.1 Precision results
The photogrammetric errors (RMSE) calculated by the Photoscan on x,y
and z-axes for the control, check and tie points of each SfM point cloud are listed
in Table 2. The point clouds show average errors of order ~3 mm on xyz on control
and check points, whereas the tie points image residual RMS was ~ 0·3 pix.
Table 2. Root mean square error (RMSE) of check points, control points and tie
points image residuals.
Plot Date
RMS tie points
image residuals
(pix)
RMSE of control
points
RMSE of check
points
(mm) (mm)
X Y Z X Y Z
1
06/06/16 0·264 2·464 2·987 2·154 2·387 3·690 1·196
22/08/16 0·236 2·164 1·626 1·901 1·135 1·583 2·690
30/11/16 0·305 2·106 2·968 1·071 1·253 2·694 1·214
22/02/17 0·295 1·572 1·450 2·477 2·100 1·613 4·850
25/05/17 0·323 2·743 3·520 1·639 1·379 3·307 2·329
28/09/17 0·271 2·755 2·431 1·540 2·605 2·742 1·472
26/04/18 0·290 1·169 0·802 0·570 1·015 1·133 1·953
2
06/06/16 0·307 3·682 2·511 2·140 2·254 3·219 3·211
22/08/16 0·292 3·752 1·827 1·120 3·328 2·255 2·765
30/11/16 0·27 3·047 1·704 1·516 3·468 1·308 3·176
22/02/17 0·275 2·864 1·907 2·299 2·905 2·427 1·803
25/05/17 0·318 3·754 2·182 2·553 1·011 1·173 1·864
73
28/09/17 0·255 2·720 1·540 1·098 2·312 1·437 2·829
26/04/18 0·388 2·421 2·013 2·274 3·080 2·066 1·834
3
06/06/16 0·363 3·507 2·015 1·964 3·017 3·797 5·500
22/08/16 0·332 3·133 2·695 1·277 3·506 1·709 0·737
30/11/16 0·277 3·068 3·563 1·261 3·441 3·826 1·399
22/02/17 0·269 2·500 2·599 1·237 1·981 2·218 2·221
25/05/17 0·331 1·721 2·193 0·999 0·943 2·298 2·658
28/09/17 0·273 2·880 1·579 1·380 2·776 2·274 1·138
26/04/18 0·292 1·544 2·280 1·172 1·463 2·701 1·646
The LoD maps show the spatial variation of precision along the plot (Fig.
3), with values ranging from 1·4 mm to 7·4 mm. The larger values were
concentrated in the area with less image overlap.
Fig. 3. Level of detection (LoD) maps showing the spatial distribution of potential
error along the flume. Changes with magnitudes smaller than the LoD can be
disregarded.
3.3.2 DEM of Difference (DoD)
74
The DoD maps obtained from the erosion plots (Fig. 4) showed
remarkable variations in relation to soil movement over the studied period, with
95% of the area showing significant changes in soil surface (i.e. larger than LoD).
Although erosion was predominant, it was also possible to detect soil deposition,
mainly in the lower part of the plots near the sediment collectors. The periods
where there were major soil movements were between November 2016 -
February 2017 and September 2017 - April 2018 (Fig. 4), which match with the
rainy season in the Southwest of Brazil. The dry season, which corresponds to
the period between May and September, was also represented by the DoD maps,
by less soil movement along the plot.
Sheet erosion was the predominant type of soil erosion over the study
period. However, between September 2017 - April 2018, it was possible to
observe the formation of rill erosion, where the highest rates of water erosion
were concentrated.
75
Fig. 4. DEM of difference (DoD) maps, overlain over hillshaded topography,
showing soil erosion over natural runoff. Colour scale ranges from red (erosion)
to blue (deposition). Transparent regions mean no significant changes (DoD is
less than the level of detection).
3.3.3 Erosion measurements
76
The soil loss values obtained by SfM showed a high correlation (R2 =
95·55%) with the traditional sediment collection method (Fig. 5). Considering the
measurements performed by both methods, it was observed that the values of
soil losses obtained through the sediment collection tended to present values
slightly higher than those found by the SfM (Table 3). However, the soil loss
measurements made by the SfM were closely related to the amount of sediments
collected in all seasons of the year, both in summer (rainy season) and winter
(dry season).
Fig. 5. The relationship between the soil loss from sediments collection and
structure from motion method. The dashed line represents the 1:1 relation. The
grey zone is the confidence interval for the mean.
77
Table 3. Averaged soil loss calculated from sediment collection and structure
from motion (SfM), and natural rainfall rates during each studied period.
Date Sediments (kg) SfM (kg) Rainfall (mm)
Jun/2016 – Aug/2016 53.04 42.57 92
Aug/2016 – Nov/2016 129.93 127.40 194
Nov/2016 – Feb/2017 418.20 338.20 661
Feb/2017 – May/2017 304.33 294.67 149
May/2017 – Sep/2017 87.13 98.33 115
Sep/2017 – Apr/2018 520.45 470.11 1121
Measurements using the SfM followed the sediment data for the three
studied plots. However, in plot 1 the magnitude of the water erosion values in
rainy periods was higher than the other two monitored plots (Fig.6).
78
Fig. 6. Soil loss measurements calculated from structure from motion (SfM) and
sediment collection for three erosion plots. Lines are to illustrate trend and do
not imply a relationship between the points. The grey bars show the amount of
rainfall during each studied period.
79
3.4 Discussion
3.4.1 Erosion measurements from SfM
This was the first time the use of SfM to determine ‘sheet’ flow has been
evaluated independently using sediment collection. The high correlation
between the soil loss from SfM and collected on runoff tanks opens up the
possibility to use SfM for erosion studies where channelized erosion is not the
principal mechanism. This represents a great step forward on soil erosion
assessment around the world. Also, due to the limitations related to erosion plots,
such as high operational costs, measurements variability, due to human
disturbance in collecting data (Zobisch et al., 1996) and plots of different sizes
(Bagarello and Ferro, 2004), the SfM can potentially increase the quality of the
global soil erosion database.
Soil erosion is a process composed of three sub-processes: erosion,
transport, and deposition (Morgan, 2005). Sediment and surface runoff
collections are restricted to the evaluation of the amount of soil lost rather than
the soil erosion volume, since the traditional method does not allow the
determination of the mass of soil moved during erosion-transport-deposition
processes. Through SfM, it is possible to generate erosion and deposition maps
that allow the volume of soil moved at different times and positions to be
determined (Fig. 4). In addition, this method can distinguish the differences
80
between soil eroded volume and soil lost volume. Also, it can be used to
investigate the sediment delivery rate (Guo et al., 2016).
However, SfM does rely on images of the soil surface, meaning that it is
not suitable for areas with significant vegetation cover. Thus, applying the SfM
technique to measure erosion in cultivated areas, where the highest soil loss
values occur, is still a major challenge. To overcome this limitation and allow soil
surface 3-D reconstruction in vegetation cover areas, vegetation filtering
algorithms are being developed (e.g. CANUPO (Brodu and Lague, 2012) or CSF
(Zhang et al., 2016)). However, the accuracy of DEMs obtained from these
techniques is not yet enough to measure laminar erosion accurately.
In periods with high precipitation values, the highest soil loss values
found by SfM, compared to the collected sediments, in plots 2 and 3 (Fig. 6),
occurred because the SfM soil loss calculations considered the variation in
microtopography and high rainfall can change the soil surface, promoting its
consolidation. Soil consolidation occurs because of gravity causing particles to
collapse due to their own weight and the impact of raindrops (Eltner et al., 2015).
SfM will also capture changes to the soil surface that are not due to erosion, for
example the consolidation of the soil following tillage (Eltner et al., 2015),
crusting and degradation of the soil structure are expected due to wetting and
drying cycles, causing reduction of soil roughness, or its disturbance by soil
animals.
3.4.2 Evaluation of SfM accuracy
81
The accuracy of the 3D point coordinates acquired from SfM can be
affected by photogrammetric factors such as image geometry and georeferencing
(James et al., 2017a). In this study, the spatial variation of LoD was related to
photogrammetry, more precisely to the image overlap along the flight. This
occurred due to the manual navigation of the UAV used in this study, which
requires operator care to achieve the necessary coverage of the monitored area.
In addition, flight speed must be adjusted to achieve the required overlap among
photographs and reduce risks of blurred images at high speeds.
Other factors that influence the accuracy of SfM models are surface types
(mainly vegetation), soil roughness, and the presence of water (Eltner et al., 2015;
James et al., 2017b). In the present study, considering only the bare soil, the SfM
results showed a strong correlation with the values obtained by the sediment
collection (Fig. 5). However, it is important to note that areas with vegetation
present a complication for the interpretation of erosion measured using UAV.
This is of great importance for regions where soil loss changes from vegetated to
a non-vegetated surface.
SfM point clouds tend to smooth the soil surface in smaller scale
roughness. This can be controlled by the quality parameters in Photoscan during
dense cloud generation; but cloud noise might increase when “ultra-high
quality” is used (Cook, 2017). Thus, care should be taken when analysing
roughness surface data by choosing flight heights, overlapping, and image
resolution to ensure accurate representation of the soil surface texture at the
desired scale. The smoothing of photogrammetric data has been observed (Smith
82
et al., 2004; Jester and Klik, 2005); however, the effect of the measurement
technique has been combined with the interpolation effect during the generation
of DEM or meshing (Lane et al., 2000).
3.5 Conclusions
This work evaluated for the first time the capacity of SfM in measuring
soil erosion comparing with independent data collected from runoff tanks. The
high correlation between the soil loss from SfM and collected on runoff tanks
opens up the possibility to use SfM for erosion studies where channelized erosion
is not the principal mechanism, enabling new insights into sheet, and interrill,
erosion processes.
The use of UAV associated with the SfM technique generates a cheap,
portable, and easy way to obtain erosion measurements on a smaller scale with
high accuracy, in contrast to the traditional standard plot methods of erosion
monitoring worldwide. The results of SfM allows not only the quantification of
soil loss, for later use in models such as USLE, but also represents the spatial and
temporal dimensions of the soil erosion process, which is of great importance in
understanding the mechanisms of the water erosion.
3.6 References
Agüera-Vega, F., Carvajal-Ramírez, F., Martínez-Carricondo, P., López, J.S-H., Mesas-
Carrascosa, F.J., García-Ferrer, A., Pérez-Porras, F.J., 2018.
83
Araya, T., Cornelis, W.M., Nyssen, J., Govaerts, B., Bauer, H., Gebreegziabher, T., Oicha,
T., Raes, D., Sayre, K.D., Haile, M., Decker, J., 2011. Effects of conservation agriculture
on runoff, soil loss and crop yield under rainfed conditions in Tigray, Northern Ethiopia.
Soil Use Manage. 27, 404–414.
Bagarello, V., Ferro, V., 2004. Plot-scale measurement of soil erosion at the experimental
area of Sparacia (southern Italy). Hydrol. Process. 18, 141–157.
Bakker, M.M., Govers, G., Rounsevell, M.D., 2004. The crop productivity–erosion
relationship: an analysis based on experimental work. Catena 57, 55–76.
Bazoffi, P., 2015. Measurement of rill erosion through a new UAV-GIS methodology.
Ital. J. Agron. 10, s1:708.
Boardman, J., Evans, R., Ford, J., 2003. Muddy floods on the South Downs, southern
England: problem and responses. Environ. Sci. Pol. 6, 69–83.
Brunier, G., Fleury, J., Anthony, E.J., Gardel, A., Dussouillez, P., 2016. Close-range
airborne structure-from-motion photogrammetry for high-resolution beach
morphometric surveys: examples from an embayed rotating beach. Geomorphology 261,
76–88.
Castillo, C., Pérez, R., James, M.R., Quinton, J.N., Taguas, E.V., Gómez, J.A., 2012.
Comparing the accuracy of several field methods for measuring gully erosion. Soil Sci.
Soc. Am. J. 76, 1319–1332.
Cerdan, O., Govers, G., Le Bissonnais, Y., Van Oost, K., Poesen, J., Saby, N., Gobin, A.,
Vacca, A., Quinton, J., Auerswald, K., Klik, A., Kwaad, F., Raclot, D., Ionita, I., Rejman,
J., Rousseva, S., Muxart, T., Roxo, M., Dostal, T., 2010. Rates and spatial variations of soil
erosion in Europe: A study based on erosion plot data. Geomorphology 122, 167–177.
Cook, K.L., 2017. An evaluation of the effectiveness of low-cost UAVs and structure from
motion for geomorphic change detection. Geomorphology 278, 195–208.
Colomina, I., Molina, P., 2014. Unmanned aerial systems for photogrammetry and
remote sensing: a review. ISPRS J. Photogramm. Remote Sens. 92, 79–97.
Di Stefano, C., Ferro, V., 2017. Testing sediment connectivity at the experimental SPA2
basin, Sicily (Italy). Land Degrad. Dev. 28, 1992–2000.
Dietrich, J.T., 2016. Riverscape mapping with helicopter-based Structure-from-Motion
photogrammetry. Geomorphology 252, 144–157.
Easa, S.M., 1988. Estimating pit excavation volume using nonlinear ground profile. J.
Surv. Eng. 114, 71–83.
Eltner, A., Baumgart, P., Maas, H.G., Faust, D., 2015. Multi-temporal UAV data for
automatic measurement of rill and interrill erosion on loess soil. Earth Surf. Process.
Landf. 40, 741–755.
84
Eltner, A., Maas, H-G., Faust, D., 2018. Soil micro-topography change detection at
hillslopes in fragile Mediterranean landscapes. Geoderma 313, 217–232.
Fang, N.F., Wang, L., Shi, Z.H. Runoff and soil erosion of field plots in a subtropical
mountainous region of China. J. Hydrol. 552, 387–395.
Fawzy, H.E-D., 2015. The accuracy of determining the volumes using close range
photogrammetry. J. Mech. Civ. Eng. 12, 10–15.
Fonstad, M.A., Dietrich, J.T., Courville, B.C., Jensen, J.L., Carbonneau, P.E., 2013.
Topographic structure from motion: a new development in photogrammetric
measurement. Earth Surf. Process. Landf. 38, 421–430.
García-Ruiz, J.M., Beguería, S., Nadal-Romero, E., González-Hidalgo, J.C., Lana-Renault,
N., Sanjuán, Y. A meta-analysis of soil erosion rates across the world. Geomorphology,
239, 160–173.
Gelman, A., Hill, J., 2007. Data analysis using regression and multilevel/hierarchical
models. Cambridge University Press, New York.
Gómez-Gutiérrez, A., Schnabel, S., Berenguer-Sempere, F., 2014. Using 3D photo-
reconstruction methods to estimate gully headcut erosion. Catena 120, 91–101.
Glendell, M., McShane, G., Farrow, L., James, M., Quinton, J., Anderson, K., Evans, M.,
Benaud, P., Rawlins, B., Morgan, D., Jones, L., Kirkham, M., DeBell, L., Quine, T., Lark,
M., Rickson, J., Brazier, R., 2017. Testing the utility of structure from motion
photogrammetry reconstructions using small unmanned aerial vehicles and ground
photography to estimate the extent of upland soil erosion. Earth Surf. Process. Landf. 42,
1860–1871.
Granshaw, S.I., 1980. Bundle adjustment methods in engineering photogrammetry.
Photogramm. Rec. 10, 181–207.
Guo, M., Shi, H., Zhao, J., Liu, P., Welbourne, D., Lin, Q., 2016. Digital close range
photogrammetry for the study of rill development at flume scale. Catena 143, 265–274.
Guo, Q., Hao, Y., Liu, B., 2015. Rates of soil erosion in China: a study based on runoff
plot data. Catena, 124, 68–76.
Hänsel, P., Schindewolf, M., Eltner, A., Kaiser, A., Schmidt, J., 2016. Feasibility of high-
resolution soil erosion measurements by means of rainfall simulations and SfM
photogrammetry. Hydrology 3 (38).
James, M.R., Quinton, J.N., 2014. Ultra-rapid topographic surveying for complex
environments: the hand-held mobile laser scanner (HMLS). Earth Surf. Process. Landf.
39, 138–142.
James, M.R., Robson, S., 2012. Straightforward reconstruction of 3D surfaces and
topography with a camera: accuracy and geoscience application. J. Geophys. Res. 117,
F03017.
85
James, M.R., Robson, S., 2014. Mitigating systematic error in topographic models derived
from UAV and ground-based image networks. Earth Surf. Process. Landf. 39, 1413–1420.
James, M.R., Robson, S., Smith, M.W., 2017a. 3-D uncertainty-based topographic change
detection with structure-from-motion photogrammetry: precision maps for ground
control and directly georeferenced surveys. Earth Surf. Process. Landf. 42, 1769–1788.
James, M.R., Robson, S., d'Oleire-Oltmanns, S., Niethammer, U., 2017b. Optimising UAV
topographic surveys processed with structure-from-motion: Ground control quality,
quantity and bundle adjustment. Geomorphology 280, 51–66.
Jester, W., Klik, A., 2005. Soil surface roughness measurement—methods, applicability,
and surface representation. Catena 64, 174–192.
Lal, R., 1998. Soil erosion impact on agronomic productivity and environment quality.
CRC Crit. Rev. Plant Sci. 17, 319–464.
Lane, S.N., James, T.D., Crowell, M.D., 2000. Application ofdigital photogrammetry to
com- plex topography for geomorphological research. Photogramm. Rec. 16, 793–821.
Lane, S.N., Westaway, R.M., Hicks, D.M., 2003. Estimation of erosion and deposition
volumes in a large, gravel-bed, braided river using synoptic remote sensing. Earth Surf.
Process. Landf. 28, 249–271.
Lowe, D., 2004. Distinctive image features from scale-invariant keypoints. Int. J. Comput.
Vis. 60, 91–110.
Mancini, F., Dubbini, M., Gattelli, M., Stecchi, F., Fabbri, S., Gabbianelli, G., 2013. Using
unmanned aerial vehicles (UAV) for high-resolution reconstruction of topography: The
structure from motion approach on coastal environments. Remote Sens. 5, 6880–6898.
Morgan, R.P.C., 2005. Soil Erosion and Conservation. third ed. Blackwell Publishing
Ltd., Cornwall.
Morgan, J.A., Brogan, D.J., Nelson, P.A., 2017. Application of Structure-from-Motion
photogrammetry in laboratory flume. Geomorphology 276, 125–143.
Neugirg, F., Stark, M., Kaiser, A., Vlacilova, M., Della Seta, M., Vergari, F., Schmidt, J.,
Becht, M., Haas, F., 2016. Erosion processes in calanchi in the upper Orcia Valley,
southern Tuscany, Italy based on multitemporal high-resolution terrestrial LiDAR and
UAV surveys. Geomorphology 269, 8–22.
Nouwakpo, S., Weltz, M., McGwire, K., 2015. Assessing the performance of structure-
from-motion photogrammetry and terrestrial lidar for reconstructing soil surface
microtopography of naturally vegetated plots. Earth Surf. Process. Landf. 41, 308–322.
Ouédraogo, M.M., Degré, A., Debouche, C., Lisein, J., 2014. The evaluation of unmanned
aerial system-based photogrammetry and terrestrial laser scanning to generate DEMs of
agricultural watersheds. Geomorphology 214, 339–355.
86
Papiernick, S.K., Schumacher, T.E., Lobb, D.A., Lindstrom, M.J., Lieser, M.L., Eynard,
A., Schumacher, J.A., 2009. Soil properties and productivity as affected by topsoil
movement within an eroded landform. Soil Tillage Res. 102, 67–77.
Phan Ha, H.A.P., Huon, S., Henry des Tureaux, T.H., Orange, D., Jouquet, P., Valentin,
C., De Rouw, A., Tran Duc, T.T., 2012. Impact of fodder cover on runoff and soil erosion
at plot scale in a cultivated catchment of North Vietnam. Geoderma 177, 8–17.
Prosdocimi, M., Burguet, M., Di Prima, S., Sofia, G., Terol, E., Rodrigo Comino, J., Cerdá,
A., Tarolli, P., 2017. Rainfall simulation and structure-from-motion photogrammetry for
the analysis of soil water erosion in Mediterranean vineyards. Sci. Total Environ. 574,
204–215.
Quinton, J.N., Govers, G., Van Oost, K., Bardgett, R.D., 2010. The impact of agricultural
soil erosion on biogeochemical cycling-s2. Nat. Geosci. 3, 1–6.
Seitz, S.M., Curless, B., Diebel, J., Scharstein, D., Szeliski, R., 2006. A comparison and
evaluation of multi-view stereo reconstruction algorithms. IEEE Conference on
Computer Vision and Pattern Recognition. IEEE Computer Society, New York.
Smith, M.J., Asal, F.F.F., Priestnall, G., 2004. The use of photogrammetry and lidar for
landscape roughness estimation in hydrodynamic studies. Int. Arch. Photogramm.
Remote. Sens. Spat. Inf. Sci. 35 (B3), 714–719.
Stöcker, C., Eltner, A., Karrasch, P., 2015. Measuring gullies by synergetic application of
UAV and close range photogrammetry — a case study from Andalusia, Spain. Catena
132, 1–11.
Stumpf, A., Malet, J.P., Kerle, N., Niethammer, U., Rothmund, S., 2013. Image-based
mapping of surface fissures for the investigation of landslide dynamics. Geomorphology
186, 12–27.
Taylor, J.R., 1997. An Introduction to Error Analysis: the Study of Uncertainties in
Physical Measurements, second edition. University Science Books: Sausalito, California.
Ullman, S., 1979. The interpretation ofstructure from motion. Proc. R. Soc. Lond. B Biol.
Sci. 203, 405–426.
Warrick, J., Ritchie, A., Adelman, G., Adelman, K., Limber, P., 2017. New techniques to
measure cliff change from historical oblique aerial photographs and structure-from-
motion photogrammetry. J. Coast. Res. 33, 39 – 55.
Westoby, M., Brasington, J., Glasser, N., Hambrey, M., Reynolds, J., 2012. ‘Structure-
from-motion’ photogrammetry: a low-cost, effective tool for geoscience applications.
Geomorphology 179, 300–314.
Westoby, M., Dunning, S., Woodward, J., Hein, A., Marrero, S., Winter, K., Sugden, D.,
2016. Interannual surface evolution of an Antarctic blue-ice moraine using multi-
temporal DEMs. Earth Surf. Dyn. 4, 515–529.
87
Wheaton, J., Brasington, J., Darby, S., Sear, D., 2010. Accounting for uncertainty in DEMs
from repeat topographic surveys: improved sediment budgets. Earth Surf. Process.
Landf. 35, 136–156.
White, J., Wulder, M., Vastaranta, M., Coops, N., Pitt, D., Woods, M., 2013. The utility of
image-based point clouds for forest inventory: a comparison with airborne laser
scanning. Forests 4, 518–536.
Zhao, J., Van Oost, K., Chen, L., Govers, G., 2016. Moderate topsoil erosion rates
constrain the magnitude of the erosion-induced carbon sink and agricultural
productivity losses on the Chinese Loess Plateau. Biogeosciences 13, 4735–4750.
Zhao, J., Yang, Z., Govers, G. Soil and water conservation measures reduce soil and
water losses in China but not down to background levels: Evidence from erosion plot
data. Catena 337, 729–741.
Zobisch, M.A., Klingspor, P., Oduor, A.R., 1996. The accuracy of manual runoff and
sediment sampling from erosion plots. J. Soil Water Conserv. 51, 231–1233.
Zuur, A.F., Ieno, E.N., Walker, N.J., Saveliev, A.A., Smith, G.M., 2009. Mixed effects
models and extensions in ecology with R. Springer, New York, New York, USA.
88
4. Evaluation of sediment source and volume of soil erosion in a gully system
using structure-from-motion and UAV data: A case study from Minas Gerais,
Brazil
4.1 Introduction
Gullies represent a significant source of sediments, especially in tropical
environments (Poesen, 2011), reaching areas of about 3·5 ha for a single gully (Lin
et al., 2015). Gully erosion can be defined as being an erosive process where the
water concentrates in the landscape, being affected by the presence of tracks and
the lack of conservation measurements in the area (Poesen et al., 2002; Valentin
et al., 2005; Lin et al., 2015). The concentrated flow reduces top soil by the gully
initiation, causing severe impacts in farmland productivity and waterways
sedimentation (Allen et al., 2018; Bastola et al., 2018; Zabihi et al., 2018).
Long-term studies report that gullies develop randomly and are linked
with the natural mass movements associated with the removal of vegetation
cover (Harvey, 1997; Lin et al., 2015). However, gully development involves
several sub-processes related to water erosion and mass movements, such as
detachment, transport and deposition of sediments, gully bank retreat, piping
and fluting (Harvey, 1992). The complex interaction between these sub-
processes, with erosion and deposition occurring simultaneously in the area
(Gómez-Gutiérrez et al., 2012), coupled with the three-dimensional nature of the
89
gullies, make it difficult to measure and quantify directly in the field (De Rose et
al., 1998; Poesen et al., 2003).
Thus, traditional methods such as pins (Desir and Marín, 2007),
microtopographic profiles (Casalí et al., 2006), surveys with total stations
(Ehirobo and Audu, 2012) and poles are being replaced by techniques based on
high-resolution photogrammetry (Castillo et al. 2012). Several studies have
quantified gully erosion through photogrammetry associated with three-
dimensional soil surface reconstruction methods, such as Structure from Motion
(SfM) (Castillo et al., 2012; Gómez-Gutiérrez et al., 2014; Kaiser et al., 2014, Di
Stefano et al., 2017; Ben Slimane et al., 2018). However, few studies used the SfM
for detailed study of sediment sources and their movement over time in the gully
environment. Considering that gullies have a complex growth dynamics, the
study of spatial and temporal evolution through aerial images are important for
the development of control strategies and mitigation of degraded areas.
Through SfM photogrammetry it is possible to elucidate better the erosive
processes that occur in the gully system, by obtaining DEMs with high spatial
and temporal resolution. With recent advances in the use and availability of
unmanned aerial vehicles (UAVs), the use of SfM photogrammetry to produce
high-resolution DEMs has become popular in geosciences (d’Oleire-Oltmanns et
al., 2012; Carollo et al., 2015; Di Stefano et al., 2017), because it is cheap, less time-
consuming, requires little knowledge due to the automation of processes and has
similar accuracy to the most accurate methods currently available (such as laser
scanning) (Castillo et al., 2012; James and Robson, 2012; Fonstad et al., 2013).
90
The knowledge of the contribution rates of rills and gully sidewalls, as
well as the quantification of sediments stored in the channels and lost from the
gully system is important for the development of effective strategies to control
soil erosion in gullies (Hosseinalizadeh et al., 2019). This spatial and temporal
variation of sediments in gully development are indicators used by land
managers to identify the stage of development and stabilization of the gully
system (Betts et al., 2003). When the amount of lost sediment becomes smaller
than that stored in the channels, it indicates a stabilization of the erosive process
in the gully (Kasai et al., 2001).
Although many studies have described the formation and development
processes of gullies (Harvey, 1992; Vandekerckhove et al., 1998; Sidorchuk et al.,
2003; Conoscenti et al., 2014), few papers have evaluated the high spatial
resolution of the dynamics of soil movement along the gully system. Using high-
resolution DEMs it is possible to elucidate the complex erosion processes that
occur simultaneously in gullies, visualize the dynamics of soil movement in the
system over time and devise effective strategies to mitigate sediment delivery in
watercourses.
The changes in the macro and micro topography of the gully system
requires the understanding of the continuous process of source-transport-
deposition of sediments (Valentin et al., 2005). Thus, the objectives of this study
were to use SfM (1) to determine the relative contribution of rills and gully
sidewalls to sediment generation, (2) to quantify the sediment volumes stored in
91
channels and lost from the gully, and (3) to quantify the total volume of
sediments produced by the gully.
4.2 Materials and Methods
4.2.1 Study area
The studied gully is located in a degraded area (Fig. 1) on the campus of
the Federal University of Lavras, Southeastern Brazil (21°13'37.3" S and
44°59'11.9" W). The study area has a humid subtropical climate and an average
annual rainfall of 1,530 mm. The gully has a total catchment area of 530 m2.
Fig. 1. Location of the study area.
4.2.2 Image acquisition for SfM
92
Images were acquired using the UAV DJI Phantom 3 Professional
integrated with a gimbal-stabilized FC300X camera with 12-megapixel (4000 ×
3000) Sony EXMOR 1/2·3 sensor, 94º field of view (FOV) and 20-mm focal length.
The lens aperture was set to f/2·8 and images acquired in RAW format. Two
flights were performed in the gully area, the first in October 2017 and the second
in May 2018.
In order to cover the complex 3-D area of the gully it was acquired oblique
images, which also added to the strength of the network geometry (James et al.,
2017a). However, as a result of the multiple camera angles, the overlap
percentage between the images was highly variable (Fig. 2a). Thus, the number
of images in which some point is present was used as the metric to describe the
image overlap. In this study, most areas were captured by more than 30
overlapping images, because of the oblique angles. Surveys comprised about 300
images, which reflects the complex nature of the gully morphology. The flying
altitudes ranged between 5 and 15 m, resulting in a nominal ground sampling
distance between 2 and 6 mm.
93
Fig. 2. Annotated computer screenshot of Photoscan showing (a) camera
positions and orientations, and (b) control point layout.
To compare the SfM results at different times, both surveys must be in the
same coordinate system. Thus, for the georeferencing, 15 permanent ground
control points (GCP) were installed in the area (Fig. 2b). The GCP coordinates
were determined by a total station (Geodetic GD2i, accuracy 2 mm), within an
arbitrary local coordinate system.
4.2.3 SfM point cloud generation
The three-dimensional point clouds (3D) were generated from the sets of
photographs using the SfM commercial software Agisoft Photoscan version 1.4.5
(Agisoft, 2018) (see Chapter 1). The photogrammetric parameters used on
Photoscan are listed in Table 1. All surface reconstructions were done through
94
cloud computing using a virtual machine with 24 cores, 128 GB RAM and two
NVIDIA Tesla K80 GPUs.
Table 1. Photoscan parameters settings used during the point cloud generation.
Point cloud: alignment parameters Setting Accuracy Highest
Generic preselection Yes Reference preselection Yes Key point limit 120,000 Tie point limit 0 Filter point by mask No
Dense point cloud: reconstruction parameters
Quality High
Depth filtering Mild
4.2.4 Change detection and 3D precision maps
To evaluate the soil surface changes in the different surveys, it was used
the precision maps (PM) variant of the Multiscale Model to Model Cloud
Compare algorithm (M3C2; Lague et al., 2013), an analytical tool implemented in
CloudCompare. M3C2-PM is a more appropriate technique for analysing
complex 3D environments than DEM of Difference (DoD) (Lague et al., 2013).
Comparisons using DEMs can overestimate errors on steep terrain since small
lateral shifts can produce large vertical differences (Cook, 2017).
The M3C2-PM algorithm finds the most appropriate normal direction for
each point and calculates the distance between the two point clouds along a
cylinder of a given radius projected along the normal. The comparisons used core
points with 1 cm spacing, a cylinder with a 30 cm diameter, and multiscale
normals with radii from 0.2 m to 1 m with a step of 0.2 m.
95
The native M3C2 uses a roughness-based metric to estimate precision, but
this is not appropriate for photogrammetric point clouds (James et al., 2017b).
Thus, in this study the PM it was used to obtain the confidence intervals in the
detection of changes between the surveys. M3C2-PM approach has a greater
capacity to detect changes in areas of complex topography, such as gullies,
considering the spatial and 3D variation of survey accuracy (James et al., 2017b).
A detailed explanation of M3C2-PM is given by James et al. (2017b).
However, in this study, the precision estimates were derived by
reprocessing the Photoscan using DBAT bundle adjustment (Murtiyoso et al.,
2018), integrated into SfM_georef (James and Robson, 2012). All these data were
provided to me by Michael James. The precision maps were generated through
the interpolation (5-mm grid size) of the vertical standard deviation (σZ) derived
by the precision estimates. It was used median as interpolation method to
minimise the influence of outliers (James et al., 2017b).
To calculate the gully erosion volume, as well as the relative contribution
of rill erosion and mass movements, the dense point clouds were interpolated (5-
mm grid size) using the Kriging method. The zones related to each type of
erosion were delimited considering as rills channels with more than 0.01 m in
width and depth (Foster, 2005) and as gullies channels of at least 0.3 m in width
and depth (Blanco and Lal, 2010). The volumes of sediments stored and lost from
the gully system were calculated using the Simpson's rule method (Easa, 1988),
which assumed non-linearity in the profile between the grid points. The volume
96
calculations and maps were performed using Surfer (Golden Software Inc.,
Golden, CO).
4.3 Results
4.3.1 Accuracy of SfM point clouds
Both surveys had similar magnitudes of photogrammetric error (Table 2).
The point clouds showed average errors of order ~ 4 mm on xyz on control and
check points, whereas the tie points image residual RMS was ~ 0·6 pix. It was
used five check points and ten control points in the area (Fig. 3).
Table 2. Root mean square error (RMSE) of check points, control points and tie
points image residuals.
Date
Number
of
Images
Dense
Cloud
Points
RMS tie
points image
residuals
(pix)
RMSE of control
points (mm)
RMSE of check
points (mm)
X Y Z X Y Z
27/10/17 277 51,002,599 0·568 4·2 2·5 1·3 4·2 4·5 2·0
26/05/18 325 65,475,214 0·561 2·8 3·3 4·5 3·2 3·5 2·6
97
Fig. 3. Location of the control and check points in the study area.
The precision maps show the spatial variation of precision on each survey
along the gully, with M3C2-PM uncertainty values ranging from 0·006 m to 0·276
m (Fig. 4). The highest values were concentrated in shaded areas and at the
bottom of the gully. The first survey was less accurate than the second one,
especially in the more complex areas.
98
Fig. 4. Precision maps for both October 2017 (A) and May 2018 (B) surveys.
The image overlap, as well as the number of images, were sufficient to
produce a distribution of the points in the clouds without large holes. For gully
erosion studies, it could be considered a large hole an empty place in the cloud
with about 10-cm spacing between the points. The 3D reconstruction of the
topography of the most complex areas of the gully was done adequately,
reproducing with fidelity the terrain morphology (Fig. 5).
99
Fig. 5. Dense point cloud showing the 3D reconstruction of complex topographic
areas inside the gully.
4.3.2 Sediment source dynamics
The significant changes found by the M3C2-PM method showed a high
visual correlation with the observed differences between both DEMs in the area
(Fig 6). Significant changes were detected in the topsoil, rill erosion and in the
mass movements, such as gully sidewalls, inside the gully.
100
Fig. 6. DEMs for the two surveys and the map showing the significant change, in
red, over the studied period.
101
During the study period, a total of 71 m3 of sediments were generated
(Table 3), and 76% of this volume was lost from the gully system. Almost all sheet
erosion was stored in the area, contributing with less than 1% to the output of
sediments from the gully. Rill erosion contributed 8% of the sediment yield in the
gully, in large part being lost in the erosion process and only 0·76 m3 stored in
the channels.
Table 3. The relative contribution of each erosion process in the gully system
between October 2017 and May 2018.
Erosion process
Sediments
generation Sediments stored Sediments lost
------------------------------- m3 ------------------------------
Sheet erosion 2·12 1·83 0·28
Rill 5·69 0·76 4·93
Gully sidewall 63·39 14·38 49·01
Total 71·20 16·97 54·22
102
Fig. 7. Erosive processes (rills and mass movements) in the gully system.
The mass movements, including gully sidewall erosion, corresponded
to 89% of the total sediments produced. However, 23% of that volume was
deposited and stored in the gully bed. Nevertheless, of the total soil loss from the
system, more than 90% was originated from the mass displacements promoted
103
by the gully sidewall, while rill erosion accounted for approximately 9% of the
sediment lost. The dynamics of the gully development, as well as the contribution
of gully side wall retreat, are well represented by the difference between the two
point clouds obtained by M3C2-PM (Fig. 7).
Fig. 8. Point cloud showing the difference (M3C2-PM distance) between the
October 2017 and May 2018 gully surveys. Colour intensity shows relative
amounts of erosion (red) and deposition (green).
104
4.4 Discussion
4.4.1 SfM measurements errors
For the study of active and dynamic environments, such as gullies, where
the variations in the soil surface are in the order of centimetres and metres, RMSE
values in the order of 4 mm for xyz, as found in this study, are acceptable. These
values are lower than those founded by Agüera-Vega et al. (2018), who also
studied topography reconstruction in complex areas using UAV. A millimetric
precision on this kind of survey is very important, because allow the assessment
of all erosion types occurring in the area, from laminar erosion to large mass
movements.
The largest photogrammetric errors, obtained in the regions of the most
complex and shaded topography (Fig. 4), can be reduced by performing flights
on cloudy days with indirect light, increasing the number of oblique images and
adding images taken in different height (Castillo et al., 2012, Gómez-Gutiérrez et
al., 2014, Stöcker et al., 2015; Carbonneau and Dietrich, 2017, James et al., 2017b).
Moreover, in areas where there is large soil movement, such as the gully
environment, it is advisable to use dGPS rather than total station (with an
arbitrary local coordinate system) to collect GCPs locations. This is to avoid
repeatable GCPs surveys due to the soil movements, especially in points located
in the bed and near the gully sidewalls.
4.4.2 Source of sediments in the gully
105
The present study showed that the gully growth occurred towards the
main erosion channels present in the area (Figs. 6 and 7). The runoff concentrated
in rills or depressions has the capacity to remove soil particles from the gully
through sluicing (Lin et al., 2015). The gully side walls usually retreat due to three
processes: mass displacement, the detachment of soil particles by splashes, or
water running along gully banks (Chaplot et al., 2011). In the studied gully, the
gully side wall retreated primarily due to the mass displacement, as showed by
the M3C2-PM distance map (Fig. 8). These results correspond to those of
Vandekerckhove et al. (2003) and Hosseinalizadeh et al. (2019).
In contrast to previous gully erosion studies (Prosser and Slade, 1994;
Inoubli et al., 2017; Ben Slimane et al., 2018), sediment generation in the studied
gully was predominantly by the mass displacement process due to the erosion of
the gully side walls. These results are similar to those found by De Rose et al.
(1998) and Betts et al. (2003). Mass movements of gully side walls are also
recognized by Harvey (2001) as an important process in the absence of extreme
rainfall events and have been related to reactivation of gullies.
Studies indicate that in stabilized gullies it is expected that the amount of
sediment stored in the channels will exceed the volume of soil lost in the gully
system (Kasai et al., 2001; Betts et al., 2003). In the present study, 54·22 m3 was
lost from the gully system in only 8 months of monitoring, a value similar to that
was found by Ben Slimane et al. (2018) for annual production of sediment in
gullies. While just 16·97m3 of sediments generated were stored on the system.
106
These results showed that the studied gully is not stabilized yet. In that
way, a detailed knowledge of the complex dynamics of gully evolution has
implications for the correct management and application of stabilization
practices of gully prone areas. The accelerated evolution of this gully
demonstrates that conservation strategies should be applied in the early stages
of the gully formation before the channels deepen and the mass movement
processes accelerate the evolution of the gully erosion. Attempts to reduce the
expansion of the gullies complex become less efficient in these advanced stages.
4.5 Conclusions
This study evaluated the relative contribution of the different erosive
processes that occur simultaneously in a gully. For the first time, the sediment
sources of a gully were quantified remotely and with millimetric precision.
Through the SfM, it was generated high resolution measurements, allowing to
evaluate even the contribution of sheet erosion in the generation of sediment of
the gully. This opens up new possibilities in the studies involving the dynamics
of gullies, since the understanding of the spatial and temporal behaviour of the
erosive processes are important in the development of control strategies and
monitoring of the evolution of a gullies complex.
The results revealed that the main source of sediment in the gully studied
was due to the mass movement processes. Rills and laminar erosions contributed
9% and 1%, respectively, to the total sediment yield, while the mass movements
107
corresponded with most of the sediment generation in the gully. Of the total
sediment produced in the system, only 24% was stored in the gully, indicating
its high activity and instability.
4.6 References
Agüera-Vega, F., Carvajal-Ramírez, F., Martínez-Carricondo, P., López, J.S-H., Mesas-
Carrascosa, F.J., García-Ferrer, A., Pérez-Porras, F.J., 2018.
Allen, P.M., Arnold, J.G., Auguste, L., White, J., Dunbar, J., 2018. Application of a simple
headcut advance model for gullies. Earth Surf. Process. Landf. 43, 202–217.
Bastola, S., Dialynas, Y.G., Bras, R.L., Noto, L.V., Istanbulluoglu, E., 2018. The role of
vegetation on gully erosion stabilization at a severely degraded landscape: a case study
from Calhoun Experimental Critical Zone Observatory. Geomorphology 308, 25–39.
Ben Slimane, A., Raclot, D., Rebai, H., Le Bissonnais, Y., Planchon, O., Bouksila, F., 2018.
Combining field monitoring and aerial imagery to evaluate the role of gully erosion in a
Mediterranean catchment (Tunisia). Catena 170, 73–83.
Betts, H.D., Trustrum, N.A., De Rose, R.C., 2003. Geomorphic changes in a complex gully
system measured from sequential digital elevation models, and implications for
management. Earth Surf. Process. Landf. 28, 1043–1058.
Blanco, H., Lal, R., 2010. Principles of soil conservation and management. Springer,
Heidelberg, Germany, 620 pp.
Carbonneau, P.E., Dietrich, J.T., 2017. Cost-effective non-metric photogrammetry from
consumer-grade sUAS: implications for direct georeferencing of structure from motion
photogrammetry. Earth Surf. Process. Landf. 42, 473–486.
Carollo, F.G., Di Stefano, C., Ferro, V., Pampalone, V., 2015. Measuring rill erosion at
plot scale by a drone-based technology. Hydrol. Process. 29, 3802–3811.
Castillo, C., Pérez, R.; James, M.R.; Quinton, J.N.; Taguas, E.V.; Gómez, J.A., 2012.
Comparing the accuracy of several field methods for measuring gully erosion. Soil Sci.
Soc. Am. J. 76, 1319–1332.
Casalí, J., Loizu, J., Campo, M.A., De Santisteban, L.M., Álvarez-Mozos, J., 2006.
Accuracy of methods for field assessment of rill and ephemeral gully erosion. Catena 67,
128–138.
Cook, K.L., 2017. An evaluation of the effectiveness of low-cost UAVs and structure from
motion for geomorphic change detection. Geomorphology 278, 195–208.
108
Conoscenti, C., Angileri, S., Cappadonia, C., Rotigliano, E., Agnesi, V., Maerker, M.,
2014. Gully erosion susceptibility assessment by means of GIS-based logistic regression:
a case of Sicily (Italy). Geomorphology 204, 399–411.
De Rose, R.C., Gomez, B., Marden, M., Trustrum, N.A., 1998. Gully erosion in Mangatu
Forest, New Zealand, estimated from digital elevation models. Earth Surf. Process.
Landf. 23, 1045–1053.
Desir, G., Marín, C., 2007. Factors controlling the erosion rates in a semi-arid zone
(Bardenas Reales, NE Spain). Catena 71, 31–40.
Di Stefano, C., Ferro, V., Palmeri, V., Pampalone, V., Agnello, F., 2017. Testing the use of
an image‐based technique to measure gully erosion at Sparacia experimental area.
Hydrol. Process. 31, 573–585.
d’Oleire-Oltmanns, S., Marzolff, I., Peter, K., Ries, J., 2012. Unmanned Aerial Vehicle
(UAV) for monitoring soil erosion in morocco. Remote Sens. 2012, 4, 3390–3416.
Easa, S.M., 1988. Estimating pit excavation volume using nonlinear ground profile. J.
Surv. Eng. 114, 71–83.
Ehiorobo, J.O., Audu, H.A.P., 2012. Monitoring of gully erosion in an urban area using
geoinformation technology. J. Emerg. Trends Eng. Appl. Sci. 3, 270–275.
Fonstad, M.A., Dietrich, J.T., Courville, B.C., Jensen, J.L., Carbonneau, P.E., 2013.
Topographic structure from motion: a new development in photogrammetric
measurement. Earth Surf. Process. Landf. 38, 421–430.
Foster, G.R., 2005. Terraces and terracing. In: Hillel, D. (Ed.), Encyclopedia of Soils in the
Environment. Academic Press, New York, pp. 135–143.
Chaplot, V., Brown, J., Dlamini, P., Eustice, T., Janeau, J.L., Jewitt, G., Lorentz, S., Martin,
L., Nontokozo-Mchunu, C., Oakes, E., Podwojewski, P., Revil, S., Rumpel, C., Zondi, N.,
2011. Rainfall simulation to identify the storm-scale mechanisms of gully bank retreat.
Agric. Water Manag. 98, 1704–1710.
CloudCompare (version 2.10) [GPL software], 2019. Retrieved from. http://www.
cloudcompare.org/.
Gómez-Gutiérrez, Á., Schnabel, S., Berenguer-Sempere, F., Lavado-Contador, F., Rubio-
Delgado, J., 2014. Using 3D photo-reconstruction methods to estimate gully headcut
erosion. Catena 120, 91–101.
Gómez-Gutiérrez, Á., Schnabel, S., De Sanjosé, J.J., Contador, F.L., 2012. Exploring the
relationships between gully erosion and hydrology in rangelands of SW Spain. Z.
Geomorphol. 56, 27–44.
Harvey, A.M., 1992. Process interactions, temporal scales and the development of
hillslope gully systems: Howgill Fells, northwest England. Geomorphology 5, 323–344.
109
Harvey, A.M., 1997. Coupling between hillslope gully systems and stream channels in
the Howgill Fells, northwest England: temporal implications. Geomorphologie 1, 3–20.
Harvey, A.M., 2001. Coupling between hillslopes and channels in upland fluvial
systems: implications for landscape sensitivity, illustrated from the Howgill Fells,
northwest England. Catena 42, 225–250.
Hosseinalizadeh, M., Kariminejad, N., Chen, W., Pourghasemi, H.R., Alinejad, M.,
Behbahani, A.M., Tiefenbacher, J.P., 2019. Spatial modelling of gully headcuts using
UAV data and four best-first decision classifier ensembles (BFTree, Bag-BFTree, RS-
BFTree, and RF-BFTree). Geomorphology 329, 184–193.
Inoubli, N., Raclot, D., Moussa, R., Habaieb, H., Le Bissonnais, Y., 2016. Soil cracking
effects on hydrological and erosive processes: a study case in Mediterranean cultivated
vertisols. Hydrol. Process. 30, 4154–4167.
James, M.R., Robson, S., 2012. Straightforward reconstruction of 3D surfaces and
topography with a camera: Accuracy and geoscience application. J. Geophys. Res. 117,
F03017.
James, M.R., Robson, S., d'Oleire-Oltmanns, S., Niethammer, U., 2017a. Optimising UAV
topographic surveys processed with structure-from-motion: Ground control quality,
quantity and bundle adjustment. Geomorphology 280, 51–66.
James, M.R., Robson, S., Smith, M.W., 2017b. 3-D uncertainty-based topographic change
detection with structure-from-motion photogrammetry: precision maps for ground
control and directly georeferenced surveys. Earth Surf. Process. Landf. 42, 1769–1788.
Kaiser, A., Neugirg, F., Rock, G., Müller, C., Haas, F., Ries, J., Schmidt, J., 2014. Small-
scale surface reconstruction and volume calculation of soil erosion in complex Moroccan
gully morphology using structure from motion. Remote Sens. 6, 7050–7080.
Kasai, M., Marutani, T., Reid, L.M., Trustrum, N.A., 2001. Estimation of temporally
averaged sediment delivery ratio using aggradational terraces in headwater catchments
of the Waipaoa River, North Island, New Zealand. Earth Surf. Process. Landf. 26,1–16.
Lague, D., Brodu, N., Leroux, J., 2013. Accurate 3D comparison of complex topography
with terrestrial laser scanner: application to the Rangitikei canyon (NZ). ISPRS J.
Photogramm. Remote Sens. 82, 10–26.
Lin, J., Huang, Y., Wang, M., Jiang, F., Zhang, X., Ge, H., 2015. Assessing the sources of
sediment transported in gully systems using a fingerprinting approach: An example
from South-east China. Catena 129, 9–17.
Lowe, D.G., 2004. Distinctive image features from scale-invariant keypoints. Int. J.
Comput. Vision 60, 91–110.
Morgan, J.A., Brogan, D.J., Nelson, P.A., 2017. Application of Structure-from-Motion
photogrammetry in laboratory flume. Geomorphology 276, 125–143.
110
Murtiyoso, A., Grussenmeyer, P., Börlin, N., Vandermeerschen, J., Freville, T., 2018.
Open Source and Independent Methods for Bundle Adjustment Assessment in Close-
Range UAV Photogrammetry. Drones, 2, 3.
Poesen, J., 2011. Challenges in gully erosion research. Landform Analysis. 17, 5–9.
Poesen, J., Nachtergaele, J., Verstraeten, G., Valentin, C., 2003. Gully erosion and
environmental change: importance and research needs. Catena 50, 91–133.
Poesen, J., Vandekerckhove, L., Nachtergaele, J., Oustwood Wijdenes, D., Verstraeten,
G., van Wesemael, B., 2002. Gully erosion in dryland environments. In: Bull, L.J., Kirkby,
M.J. (Eds.), Dryland Rivers: Hydrology and Geomorphology of Semi-arid Channels.
John Wiley & Son Ltd., pp. 229–262.
Prosser, I.P., Slade, C.J., 1994. Gully formation and the role of valley-floor vegetation,
southeastern Australia. Geology 22, 1127–1130.
Seiz, S.M., Curless, B., Diebel, J., Scharstein, D., Szeliski, R., 2006. A Comparison and
Evaluation of Multi-View Stereo Reconstruction Algorithms. IEEE Conference on
Computer Vision ad Pattern Recognition. IEEE Computer Society, New York.
Sidorchuk, A., Marker, M., Moretti, S., Rodolfi, G., 2003. Gully erosion modelling and
landscape response in the Mbuluzi River catchment of Swaziland. Catena 50, 507–525.
Stöcker, C., Eltner, A., Karrasch, P., 2015. Measuring gullies by synergetic application of
UAV and close range photogrammetry – A case study from Andalusia, Spain. Catena
132,1–11.
Valentin, C., Poesen, J., Li, Y., 2005. Gully erosion: impacts, factors and control. Catena
63, 132–153.
Vandekerckhove, L., Poesen, J., Govers, G., 2003. Medium-term gully headcut retreat
rates in Southeast Spain determined from aerial photographs and ground
measurements. Catena 50, 329–352.
Vandekerckhove, L., Poesen, J., Wijdenes, D.O., de Figueiredo, T., 1998. Topographical
thresholds for ephemeral gully initiation in intensively cultivated areas of the
Mediterranean. Catena 33, 271–292.
Zabihi, M., Pourghasemi, H.R., Motevalli, A., Zakeri, M.A., 2019. Gully erosion
modeling using GIS-based data mining techniques in Northern Iran: a comparison
between boosted regression tree and multivariate adaptive regression spline. In:
Pourghasemi, H., Rossi, M. (Eds.), Natural Hazards GIS-based Spatial Modeling Using
Data Mining Techniques. Advances in Natural and Technological Hazards Re- search.
vol. 48. Springer, Cham.
111
5. General Conclusions
This thesis contributes to a new insight on studies of erosive processes in
the Earth’s surface, by investigating the efficiency and accuracy of digital
photogrammetry in a detailed study of the main types and scales of occurrence
of water erosion.
Considering that soil erosion is composed of three sub-processes that
occur on the soil surface: erosion, transport and deposition. Collecting sediment
and runoff allows only the assessment of soil volume lost, not soil erosion
volume. Thus, this thesis presents techniques that allow the visualization,
spatialization and quantification of soil movement during erosion-transport-
deposition processes, allowing to distinguish the differences between soil erosion
volume and lost soil volume, besides investigating the relative contribution. of
each type of erosion in the monitored area. The erosion map, represented by the
erosive processes in the area, is a powerful tool in the elaboration of effective soil
and water conservation management strategies.
Under laboratory and controlled conditions, the digital close-range
photogrammetry provided a detailed study of the development of rill erosion,
allowing the extraction of morphological indicators and quantification of soil
movement during runoff with millimetric precision. In addition, it was possible
to correlate water flow rate and slope gradient with the rill erosion development.
It was also evaluated the use of SfM combined with UAV images, in
obtaining soil loss measurements in areas where channelized erosion is not the
112
main mechanism. This represents the first time this kind of research was
conducted, providing valuable insight into a new direction for soil erosion
studies in erosion plots. The results of the soil losses obtained by UAV-SfM
presented a high correlation with the sediments collected in the plots. This is of
great relevance in the context of the monitoring and modelling of water erosion,
since the quantification of soil loss around the world is mainly done using plots,
a method that presents high operational cost. In addition, the study of laminar
erosion through the UAV-SfM allows not only to calculate the soil loss values but
to visualize the spatial variation of the erosion process (detachment, transport
and deposition) practically in real time along the area.
Finally, the application of the UAV-SfM technique in gully erosion was
evaluated. For the first time, a study was carried out evaluating the relative
contribution of the different types of erosion (sheet, rill and gully sidewall) in the
gully development. This was possible due to the millimetric level of precision of
the point clouds, allowing to evaluate even the contribution of the laminar
erosion, which is new in gullies studies. As a result, it was possible to quantify
sediments volumes stored in the channels and lost from the gully system, as well
as to determine the main sediment sources. Since gully development studies are
quite difficult to perform on field, due to the 3-D nature and the several factors
acting simultaneously, the UAV-SfM proved to be effective in the gully
monitoring and could be an important tool in the development of strategies of
control and mitigation of this serious environmental problem that affects several
agricultural and urban areas in tropical regions.
113
In summary, the results presented here indicate that the use of ground and
air-based photogrammetry are precise tools in detecting soil surface changes and
can be used to assess water erosion in its various forms of occurrence in nature.
In addition, the UAV-SfM has proven to be a very useful technique for
monitoring erosion over time, especially in hard-to-reach areas.
6. Future Work
This thesis advances our understanding of the use of ground and air-based
photogrammetry for soil erosion assessment. However, it also highlighted
several challenges, that would provide interesting future work.
• The results demonstrate that on steep slopes the soil erosion is
dominated by the rill erosion with less rill network density. It
would be interesting to develop a time-lapse study in order to see
the rill erosion evolution in real time. To make it possible, it is
necessary to think about an experimental design to avoid the water
reflection during the 3-D reconstruction of the soil surface.
• The results of the soil losses obtained by the UAV-SfM technique
presented high accuracy when compared to the sediments collected
from the erosion plots. However, it would be interesting for future
studies to evaluate the effectiveness of UAV-SfM in water erosion
monitoring with the presence of vegetation and different levels of
114
soil cover. For this, the accuracy of the mathematical algorithms
that filter the point cloud vegetation in order to estimate the
covered soil surface should be tested.
• The 3-D reconstruction of the studied gully presented millimetric
precision. However, due to the complexity of the soil surface in the
gully, shading areas present a lower density of points, affecting the
level of detection in these regions. Thus, it would be interesting for
future studies to combine UAV and terrestrial images, using
consumer-grade or smartphones cameras, in gully modelling. In
addition, in large gullies, where it is difficult to install control
points, it would be interesting to test the effectiveness of UAV
coupled with post-processed kinematic (PPK) or RTK GPS in the
generation of DEMs and point clouds.
115
Appendix A
ANOVA results for the effects of slope gradient and water flow rate on rill
erosion and soil loss (Chapter 2).
Factors df Sum Sq Mean Sq F value P (> F)
Rill erosion
Slope 1 1·3488 1·3488 161·5766 4·32 × 10-6 ***
Water flow 1 10·3549 10·3549 1240·4187 3·86 × 10-9 ***
Slope × Water flow 1 0·0486 0·0486 5·8185 0·047 *
Residuals 7 0·0584 0·0083
Soil loss
Slope 1 0·0362 0·0362 1·5714 0·2502
Water flow 1 5·4430 5·4430 236·0131 1·19 × 10-6 ***
Slope × Water flow 1 0·4785 0·4785 20·7468 0·002621 **
Residuals 7 0·1614 0·0231 Significance: *** P< 0·001, ** P<0·01, * P<0·05